ALBERT

Description: 〈p〉Publication date: August 2019〈/p〉 〈p〉〈b〉Source:〈/b〉 ISPRS Journal of Photogrammetry and Remote Sensing, Volume 154〈/p〉 〈p〉Author(s): 〈/p〉

Description: 〈p〉Publication date: September 2019〈/p〉 〈p〉〈b〉Source:〈/b〉 ISPRS Journal of Photogrammetry and Remote Sensing, Volume 155〈/p〉 〈p〉Author(s): Sheng Wang, Andreas Baum, Pablo J. Zarco-Tejada, Carsten Dam-Hansen, Anders Thorseth, Peter Bauer-Gottwein, Filippo Bandini, Monica Garcia〈/p〉 〈div xml:lang="en"〉 〈h5〉Abstract〈/h5〉 〈div〉〈p〉Unlike satellite earth observation, multispectral images acquired by Unmanned Aerial Systems (UAS) provide great opportunities to monitor land surface conditions also in cloudy or overcast weather conditions. This is especially relevant for high latitudes where overcast and cloudy days are common. However, multispectral imagery acquired by miniaturized UAS sensors under such conditions tend to present low brightness and dynamic ranges, and high noise levels. Additionally, cloud shadows over space (within one image) and time (across images) are frequent in UAS imagery collected under variable irradiance and result in sensor radiance changes unrelated to the biophysical dynamics at the surface. To exploit the potential of UAS for vegetation mapping, this study proposes methods to obtain robust and repeatable reflectance time series under variable and low irradiance conditions. To improve sensor sensitivity to low irradiance, a radiometric pixel-wise calibration was conducted with a six-channel multispectral camera (mini-MCA6, Tetracam) using an integrating sphere simulating the varying low illumination typical of outdoor conditions at 55〈sup〉o〈/sup〉N latitude. The sensor sensitivity was increased by using individual settings for independent channels, obtaining higher signal-to-noise ratios compared to the uniform setting for all image channels. To remove cloud shadows, a multivariate statistical procedure, Tucker tensor decomposition, was applied to reconstruct images using a four-way factorization scheme that takes advantage of spatial, spectral and temporal information simultaneously. The comparison between reconstructed (with Tucker) and original images showed an improvement in cloud shadow removal. Outdoor vicarious reflectance validation showed that with these methods, the multispectral imagery can provide reliable reflectance at sunny conditions with root mean square deviations of around 3%. The proposed methods could be useful for operational multispectral mapping with UAS under low and variable irradiance weather conditions as those prevalent in northern latitudes.〈/p〉〈/div〉 〈/div〉

Description: 〈p〉Publication date: September 2019〈/p〉 〈p〉〈b〉Source:〈/b〉 ISPRS Journal of Photogrammetry and Remote Sensing, Volume 155〈/p〉 〈p〉Author(s): Peifeng Ma, Tao Li, Chaoyang Fang, Hui Lin〈/p〉 〈div xml:lang="en"〉 〈h5〉Abstract〈/h5〉 〈div〉〈p〉This paper analyzed the dynamic behaviors of a high-speed railway (HSR) bridge and evaluated the possibility of measuring the sub-millimeter settlement and uplift using COSMO-SkyMed images. The capability of synthetic aperture radar interferometry (InSAR) technologies with the purpose of sub-millimeter deformation monitoring without ground control points has rarely been studied. In this paper, we conducted a tentative test for measuring the sub-millimeter settlement and uplift of a HSR bridge. Persistent scatterer (PS) and distributed scatterer (DS) were jointly detected to increase the point density. The temperature model was introduced to separate the thermal expansion and linear deformation. By analyzing the HSR structure and time-series deformation, we infer that PS points correspond to double-bounce scatterers mainly generated by the interactions between the girder and track slab and between the girder and fender, and DS points correspond to single-bounce scatterers generated by the bridge girder surface. The accuracy of linear deformation velocity and time-series deformation were evaluated, respectively. Under the assumptions by qualitative analysis, the results demonstrate that COSMO-SkyMed is capable of achieving sub-millimeter accuracy in linear deformation velocity. However, the leveling validation implies that it is difficult to achieve sub-millimeter accuracy in time-series deformation because of the uncertainties from incorrect ground data, location and time difference between the InSAR and ground data, the presence of lateral deformation, improper removal of the atmospheric phase screen, and inconsistency between the air temperature and bridge temperature.〈/p〉〈/div〉 〈/div〉

Description: 〈p〉Publication date: August 2019〈/p〉 〈p〉〈b〉Source:〈/b〉 ISPRS Journal of Photogrammetry and Remote Sensing, Volume 154〈/p〉 〈p〉Author(s): Hao Fang, Florent Lafarge〈/p〉 〈div xml:lang="en"〉 〈h5〉Abstract〈/h5〉 〈div〉〈p〉Analyzing and extracting geometric features from 3D data is a fundamental step in 3D scene understanding. Recent works demonstrated that deep learning architectures can operate directly on raw point clouds, i.e. without the use of intermediate grid-like structures. These architectures are however not designed to encode contextual information in-between objects efficiently. Inspired by a global feature aggregation algorithm designed for images (Zhao et al., 2017), we propose a 3D pyramid module to enrich pointwise features with multi-scale contextual information. Our module can be easily coupled with 3D semantic segmentation methods operating on 3D point clouds. We evaluated our method on three large scale datasets with four baseline models. Experimental results show that the use of enriched features brings significant improvements to the semantic segmentation of indoor and outdoor scenes.〈/p〉〈/div〉 〈/div〉

Description: 〈p〉Publication date: September 2019〈/p〉 〈p〉〈b〉Source:〈/b〉 ISPRS Journal of Photogrammetry and Remote Sensing, Volume 155〈/p〉 〈p〉Author(s): Deepak Gautam, Arko Lucieer, Christopher Watson, Colin McCoull〈/p〉 〈div xml:lang="en"〉 〈h5〉Abstract〈/h5〉 〈div〉〈p〉This study addresses the correction of lever-arm offset and boresight angle, and field of view (FOV) determination to enable accurate footprint determination of a spectroradiometer mounted on an unmanned aircraft system (UAS). To characterise the footprint, an accurate determination of the spectroradiometer position and orientation (pose) must be acquired with a global navigation satellite system (GNSS) and an inertial measurement unit (IMU). Accurate pose estimation requires an accurate lever-arm and boresight correction between the pose measuring sensors and the spectroradiometer. Similarly, the spectroradiometer FOV is required to determine the footprint size as a function of above ground level (AGL) flying height. The system used in this study consists of an IMU with dual-frequency and dual-antenna GNSS receiver, a machine vision camera, and a point-measuring spectroradiometer (Ocean Optics QE Pro). The lever-arm offset was determined from a scaled 3D point cloud of the system, created using photos of the airframe and processed with the structure-from-motion (SfM) algorithm. The boresight angles were estimated with stationary experiments by computing the difference between the orientations of the IMU, the spectroradiometer, and the camera. The orientation of the spectroradiometer was determined by moving a spectrally distinct target into the FOV. The orientation of IMU was measured by averaging its readings during the stationary epoch, while SfM was employed as an independent technique to estimate the orientation of the camera. The footprint of the spectroradiometer for a combination of AGL height and Gershun tube aperture ring was determined experimentally, enabling computation of the effective FOV. In-flight validation of the lever-arm and boresight correction was performed by comparing the corrected pose of the co-mounted camera with the pose derived from SfM as the reference. Our experimental results demonstrate that controlled determination and correction of lever-arm and boresight increases the pose estimation accuracy and thereby supports the direct georeferencing of a UAS-mounted spectroradiometer point observation.〈/p〉〈/div〉 〈/div〉

Description: 〈p〉Publication date: September 2019〈/p〉 〈p〉〈b〉Source:〈/b〉 ISPRS Journal of Photogrammetry and Remote Sensing, Volume 155〈/p〉 〈p〉Author(s): Ruixi Zhu, Li Yan, Nan Mo, Yi Liu〈/p〉 〈div xml:lang="en"〉 〈h5〉Abstract〈/h5〉 〈div〉〈p〉Supervised scene classification of aerial images is of great importance in land-cover classification. However, annotating the labeled data required for the conventional classifiers and convolutional neural networks (CNNs), costs much manpower and time. Domain adaptation methods can overcome this problem, to some extent, by transferring previously labeled data to the new images, but the classification models trained from the previously labeled data are not discriminative enough for classifying aerial images from other domains because of the data distribution differences caused by the variations in sensors, natural environments, seasons, angles, locations, and so on. In order to solve this problem, we propose a semi-supervised center-based discriminative adversarial learning (SCDAL) framework integrating three parts, namely filtering out easy triplets, proposed hard triplet loss, and the adversarial learning with center loss. In the SCDAL framework, a difficulty measure is proposed to remove easy triplets under the constraint of between-class dissimilarity and intra-class similarity and better distinguish hard triplets. The filtered triplets are then used to train a more discriminative source feature extractor with the proposed hard triplet loss combining the hardest triplet loss and semi-hard triplet loss. Adversarial learning with center loss is also proposed to reduce the feature distribution bias between the source and target feature extractors and increase the discriminative ability of the target feature extractor. The SCDAL framework is tested on two large aerial images as a case study. The experimental results demonstrate that when adequate previously labeled data but limited labeled target data exist, the SCDAL framework is superior to most of the existing domain adaptation methods, with an improvement of at least 3% in overall accuracy. It is also proved that removing easy triplets, proposed hard triplet loss, and the adversarial learning with center loss all help to improve the overall accuracy.〈/p〉〈/div〉 〈/div〉

Description: 〈p〉Publication date: October 2019〈/p〉 〈p〉〈b〉Source:〈/b〉 ISPRS Journal of Photogrammetry and Remote Sensing, Volume 156〈/p〉 〈p〉Author(s): Chenglu Wen, Changbin You, Hai Wu, Cheng Wang, Xiaoliang Fan, Jonathan Li〈/p〉 〈div xml:lang="en"〉 〈h5〉Abstract〈/h5〉 〈div〉〈p〉The mapping of road boundaries provides critical information about roads for urban road traffic safety. This paper presents a deep learning-based framework for recovering 3D road boundary using multi-source data, which include mobile laser scanning (MLS) point clouds, spatial trajectory data, and remote sensing images. The proposed road recovery method uses extracted 3D road boundaries from MLS point clouds as inputs. First, after automatic erroneous boundary removal, a CNN-based boundary completion model completes road boundaries. Then, to refine the imperfect road boundaries, road centerlines generated from dynamic taxi GPS trajectory data and remote sensing images are used as completion guidance for a generative adversarial nets model to obtain more accurate and complete road boundaries. Finally, after associating a sequence of taxi GPS recorded trajectory points with the correct 3D road boundaries, inherent geometric road characteristics and road dynamic information are extracted from the complete boundaries and taxi GPS trajectory data, respectively. The testing dataset contains two urban road MLS datasets, and the KITTI dataset. The experimental results on point clouds from different sensors demonstrate our proposed method is effective and promising for recovering 3D road boundary and extracting road characteristics.〈/p〉〈/div〉 〈/div〉

Description: 〈p〉Publication date: October 2019〈/p〉 〈p〉〈b〉Source:〈/b〉 ISPRS Journal of Photogrammetry and Remote Sensing, Volume 156〈/p〉 〈p〉Author(s): Yonghong Hu, Meiting Hou, Gensuo Jia, Chunlei Zhao, Xiaoju Zhen, Yanhua Xu〈/p〉 〈div xml:lang="en"〉 〈h5〉Abstract〈/h5〉 〈div〉〈p〉Urban heat island (UHI) can be characterized and quantified to understand the modification of urban surfaces on the local and regional climate. This study examines UHI variation across three megacities that are located in rapid urbanization regions in eastern China (Beijing, Shanghai and Guangzhou). These cities are located within a warm temperate climate zone, north subtropical climate zone, and lower subtropical climate zone, respectively. Satellite-based land surface temperature (LST) data and air temperature records from 2003 to 2016 were used to identify surface urban heat island (SUHI) and canopy urban heat island (CUHI), respectively. Generally, the average annual SUHI is higher than the CUHI, with the greatest UHIs appearing in Beijing (SUHI: 2.33 ± 0.18 °C, CUHI: 1.45 ± 0.54 °C). UHI changes across latitudes were negatively related to humidity variation, with higher UHI in drier climates. Seasonal UHI analysis suggests that a lower SUHI would occur in winter and a higher UHI in spring and summer, except for Guangzhou. CUHI in dry season was higher than in wet season for all three megacities, and the largest CUHI (2.10 ± 0.33 °C) appeared in winter in Beijing. Various patterns of seasonal cycles of SUHI and CUHI were related to monthly precipitation and solar insolation. Annual average daytime SUHI was higher than the nighttime SUHI, and larger daytime SUHI appeared in Guangzhou, contrasting with Shanghai and Beijing. The difference between SUHI and CUHI for all seasons was also high in Guangzhou. UHI changes were considered to be altered by warm and wet conditions in mega-cities of eastern China, and heat transportation from urban surface to urban canopy provided some possible understanding of the UHI change.〈/p〉〈/div〉 〈/div〉

Description: 〈p〉Publication date: October 2019〈/p〉 〈p〉〈b〉Source:〈/b〉 ISPRS Journal of Photogrammetry and Remote Sensing, Volume 156〈/p〉 〈p〉Author(s): Ping Lu, Shibiao Bai, Veronica Tofani, Nicola Casagli〈/p〉 〈div xml:lang="en"〉 〈h5〉Abstract〈/h5〉 〈div〉〈p〉Long-term InSAR techniques, such as Persistent Scatterer Interferometry and Distributed Scatterer Interferometry, are effective approaches able to detect slow-moving landslides with millimeter precision. This study presents a novel approach of optimized hot spot analysis (OHSA) on persistent scatterers (PS) and distributed scatterers (DS), and evaluates its performance on detection of landslides across the Volterra area in central Tuscany region of Italy. 1625 ascending and 2536 descending PS processed from eight years (2003–2010) of ENVISAT images were produced by the PS-InSAR technique. In addition, 16,493 ascending and 9746 descending PS/DS measurement points (MP) processed from four years (2011–2014 for ascending orbits and 2010–2013 for descending orbits) of COSMO-SkyMed images were collected by the SqueeSAR approach. The OHSA approach was then implemented on the derived PS and DS through the analysis of incremental spatial autocorrelation and the Getis-Ord 〈em〉G〈/em〉〈sub〉〈em〉i〈/em〉〈/sub〉* statistics. As a result of OHSA, PS and DS MP that are statistically significant with velocity 〈math xmlns:mml="http://www.w3.org/1998/Math/MathML" altimg="si1.svg"〉〈mrow〉〈mo〉〉〈/mo〉〈/mrow〉〈/math〉〈math xmlns:mml="http://www.w3.org/1998/Math/MathML" altimg="si2.svg"〉〈mrow〉〈mo stretchy="false"〉|〈/mo〉〈mo〉±〈/mo〉〈/mrow〉〈/math〉2〈math xmlns:mml="http://www.w3.org/1998/Math/MathML" altimg="si3.svg"〉〈mrow〉〈mo stretchy="false"〉|〈/mo〉〈/mrow〉〈/math〉 mm/year, 〈em〉p〈/em〉-value 〈math xmlns:mml="http://www.w3.org/1998/Math/MathML" altimg="si4.svg"〉〈mrow〉〈mo〉〈〈/mo〉〈/mrow〉〈/math〉 0.01 and 〈em〉z〈/em〉-score 〈math xmlns:mml="http://www.w3.org/1998/Math/MathML" altimg="si5.svg"〉〈mrow〉〈mo〉〉〈/mo〉〈/mrow〉〈/math〉〈math xmlns:mml="http://www.w3.org/1998/Math/MathML" altimg="si6.svg"〉〈mrow〉〈mo stretchy="false"〉|〈/mo〉〈mo〉±〈/mo〉〈/mrow〉〈/math〉2.58〈math xmlns:mml="http://www.w3.org/1998/Math/MathML" altimg="si7.svg"〉〈mrow〉〈mo stretchy="false"〉|〈/mo〉〈/mrow〉〈/math〉 were recognized as hot spots (HS). Meanwhile, a landslide inventory covering the Volterra area was manually prepared as the reference data for accuracy assessment of landslide detection. The results indicate that, in terms of OHSA-derived ENVISAT HS, the detection accuracy can be improved from 23.3% to 25.3% and from 50.7% to 66.4%, with decreased redundancy from 5.3% to 3.7% and from 5.3% to 2.4%, for ascending and descending orbits, respectively. In addition, for OHSA-derived Cosmo-SkyMed HS, the detection accuracy can be improved from 57.7% to 70.3% and from 73.8% to 81.5%, with decreased redundancy from 3.1% to 1.7% and from 3.4% to 2.1%, for ascending and descending orbits, respectively. Compared to traditional HS analysis such as Persistent Scatterers Interferometry Hot Spot and Cluster Analysis (PSI-HCA), OHSA has the significant advantage that the scale distance used for the Getis-Ord 〈em〉G〈/em〉〈sub〉〈em〉i〈/em〉〈/sub〉* statistics can be automatically determined by the analysis of incremental spatial autocorrelation and accordingly no manual intervention or additional digital terrain model (DTM) is further needed. The proposed method is very succinct and can be easily implemented in diverse geographic information system (GIS) platforms. To the best of our knowledge, this is the first time that OHSA has been applied to PS and DS.〈/p〉〈/div〉 〈/div〉

Description: 〈p〉Publication date: October 2019〈/p〉 〈p〉〈b〉Source:〈/b〉 ISPRS Journal of Photogrammetry and Remote Sensing, Volume 156〈/p〉 〈p〉Author(s): Hui Cao, Pengjie Tao, Haihong Li, Jun Shi〈/p〉 〈div xml:lang="en"〉 〈h5〉Abstract〈/h5〉 〈div〉〈p〉High-precision georeferencing of satellite stereo images becomes a key issue in multi-source data integration, geometric processing, and analysis for many remote sensing applications. Using existing geographic data, including readily available public data, as geometric reference for georeferencing satellite imagery is an effective and feasible way to improve the positional accuracy and reduce the costs and workforce restrictions on ground control points (GCPs) acquisition. This study proposes a novel bundle adjustment approach of push broom satellite imagery based on an equivalent geometric sensor model (EGSM). As an extension of the collinearity equations, EGSM is equivalent to the rigorous geometric sensor model, whose interior and exterior orientation parameters have clear geometric interpretations. The initial values of EGSM’s parameters can be completely recovered from the RPCs and optimized within the block adjustment without using any metadata of the satellite imagery. Furthermore, a publicly accessible digital elevation model (DEM) is used as constraints in the bundle adjustment to improve the direct georeferencing accuracies without GCPs. A set of pseudo observation equations is incorporated into the EGSM associated with appropriate weights based on the estimated variance of height differences between the tie points and the DEM surface. The performance of the proposed approach was evaluated by using 143 stereo scenes of Chinese Ziyuan-3 (ZY-3) images and the global DEM data SRTM GL3, ASTER GDEM, and AW3D30. The experimental results revealed that the proposed approach could improve not only vertical accuracies but also horizontal accuracies of satellite stereo images. The geopositional accuracy after adjustment is dependent on the quality and accuracy of the reference DEM itself. With AW3D30 as exclusive controls, the horizontal and vertical root mean square errors (RMSEs) of the experimental images are reduced from 17.3 m and 2.6 m to 2.5 m and 1.5 m, respectively. The horizontal accuracy corresponds to about 1.2 pixels in the image space, and the vertical accuracy is about 0.6 pixels with the base-height ratio of ZY-3 stereo images being 0.81. The generazation ability of the approach was further proved by the experiments of WorldView-3 stereo images.〈/p〉〈/div〉 〈/div〉

Description: 〈p〉Publication date: October 2019〈/p〉 〈p〉〈b〉Source:〈/b〉 ISPRS Journal of Photogrammetry and Remote Sensing, Volume 156〈/p〉 〈p〉Author(s): Hainan Cui, Shuhan Shen, Wei Gao, Hongmin Liu, Zhiheng Wang〈/p〉 〈div xml:lang="en"〉 〈h5〉Abstract〈/h5〉 〈div〉〈p〉Incremental Structure-from-Motion (SfM) techniques have exhibited superior practicability in many recent studies; however, efficiency and robustness remain key challenges for these techniques. In this work, we propose a new incremental SfM method that overcomes these problems in a united framework that contains two iteration loops. The inner loop is a track selection loop, where a well-conditioned subset of the feature tracks is iteratively selected to accelerate the time-consuming bundle adjustment. The outer loop is a camera registration loop, where the a priori camera rotations are estimated via rotation averaging on multiple orthogonal maximum spanning trees (OMSTs) of the view-graph and used as weak supervision for the registration. The calibrated camera poses that agree with the a priori camera rotations are preferentially registered, and after all the consistent cameras have been calibrated, the remaining cameras are incrementally registered. The results of extensive experiments demonstrate that our system can reconstruct both general and ambiguous image datasets, and our system outperforms many state-of-the-art SfM systems in terms of efficiency and robustness.〈/p〉〈/div〉 〈/div〉

Description: 〈p〉Publication date: November 2019〈/p〉 〈p〉〈b〉Source:〈/b〉 ISPRS Journal of Photogrammetry and Remote Sensing, Volume 157〈/p〉 〈p〉Author(s): Yuan Wang, Qiangqiang Yuan, Tongwen Li, Huanfeng Shen, Li Zheng, Liangpei Zhang〈/p〉 〈div xml:lang="en"〉 〈h5〉Abstract〈/h5〉 〈div〉〈p〉Aerosol optical depth (AOD) is a pivotal parameter to reflect aerosol properties, such as aerosol radiative forcing and atmospheric corrections of the aerosol effect. Unfortunately, the valid pixels of moderate resolution imaging spectroradiometer (MODIS) AOD products are scarce, which has attracted great attention from scholars. In recent years, numerous AOD recovering algorithms have been proposed and the algorithms merely employing a single temporal AOD image are regarded as the most convenient and flexible for large-scale practical applications. However, current algorithms face the challenge of insufficiently considering the impacts of aerosol variation resulted from the temporal difference. Meanwhile, the improvement of AOD valid pixels is also poor due to the scarce excavation of complementary information. In order to address these issues, a novel algorithm of spatial-temporal hybrid fusion considering aerosol variation mitigation (ST-AVM) is developed to fill the missing pixels in Aqua AOD products with a single Terra AOD image in large scale. The results show that the total recovered AOD products nearly maintain the original accuracy of MODIS. Meanwhile, the AOD coverage is significantly improved in the study areas and the degrees of improvements regionally vary. Overall, the AOD coverage over land is increased by 123.9% (from 20.5% to 45.9%) after the recovery. Besides, the spatial distribution of recovered monthly AOD products remains fairly consistent as the original Aqua. Also, the recovered annual AOD spatial distribution shows more coherent, which indicates the reliability of ST-AVM algorithm.〈/p〉〈/div〉 〈/div〉

Description: 〈p〉Publication date: October 2019〈/p〉 〈p〉〈b〉Source:〈/b〉 ISPRS Journal of Photogrammetry and Remote Sensing, Volume 156〈/p〉 〈p〉Author(s): Jean-Daniel Sylvain, Guillaume Drolet, Nicolas Brown〈/p〉 〈div xml:lang="en"〉 〈h5〉Abstract〈/h5〉 〈div〉〈p〉Tree mortality is an important forest ecosystem variable having uses in many applications such as forest health assessment, modelling stand dynamics and productivity, or planning wood harvesting operations. Because tree mortality is a spatially and temporally erratic process, rates and spatial patterns of tree mortality are difficult to estimate with traditional inventory methods. Remote sensing imagery has the potential to detect tree mortality at spatial scales required for accurately characterizing this process (e.g., landscape, region). Many efforts have been made in this sense, mostly using pixel- or object-based methods. In this study, we explored the potential of deep Convolutional Neural Networks (CNNs) to detect and map tree health status and functional type over entire regions. To do this, we built a database of around 290,000 photo-interpreted trees that served to extract and label image windows from 20 cm-resolution digital aerial images, for use in CNN training and evaluation. In this process, we also evaluated the effect of window size and spectral channel selection on classification accuracy, and we assessed if multiple realizations of a CNN, generated using different weight initializations, can be aggregated to provide more robust predictions. Finally, we extended our model with 5 additional classes to account for the diversity of landcovers found in our study area. When predicting tree health status only (live or dead), we obtained test accuracies of up to 94%, and up to 86% when predicting functional type only (broadleaf or needleleaf). Channel selection had a limited impact on overall classification accuracy, while window size increased the ability of the CNNs to predict plant functional type. The aggregation of multiple realizations of a CNN allowed us to avoid the selection of suboptimal models and help to remove much of the speckle effect when predicting on new aerial images. Test accuracies of plant functional type and health status were not affected in the extended model and were all above 95% for the 5 extra classes. Our results demonstrate the robustness of the CNN for between-scene variations in aerial photography and also suggest that this approach can be applied at operational level to map tree mortality across extensive territories.〈/p〉〈/div〉 〈/div〉

Description: 〈p〉Publication date: October 2019〈/p〉 〈p〉〈b〉Source:〈/b〉 ISPRS Journal of Photogrammetry and Remote Sensing, Volume 156〈/p〉 〈p〉Author(s): Kai Yue, Lei Yang, Ruirui Li, Wei Hu, Fan Zhang, Wei Li〈/p〉 〈div xml:lang="en"〉 〈h5〉Abstract〈/h5〉 〈div〉〈p〉Fine-grained semantic segmentation results are typically difficult to obtain for subdecimeter aerial imagery segmentation as a result of complex remote sensing content and optical conditions. Recently, convolutional neural networks (CNNs) have shown outstanding performance on this task. Although many deep neural network structures and techniques have been applied to improve accuracy, few have attended to improving the differentiation of easily confused classes. In this paper, we propose TreeUNet, a tool that uses an adaptive network to increase the classification rate at the pixel level. Specifically, based on a deep semantic model infrastructure, a Tree-CNN block in which each node represents a ResNeXt unit is constructed adaptively in accordance with the confusion matrix and the proposed TreeCutting algorithm. By transmitting feature maps through concatenating connections, the Tree-CNN block fuses multiscale features and learns best weights for the model. In experiments on the ISPRS two-dimensional Vaihingen and Potsdam semantic labelling datasets, the results obtained by TreeUNet are competitive among published state-of-the-art methods. Detailed comparison and analysis show that the improvement brought by the adaptive Tree-CNN block is significant.〈/p〉〈/div〉 〈/div〉

Description: 〈p〉Publication date: October 2019〈/p〉 〈p〉〈b〉Source:〈/b〉 ISPRS Journal of Photogrammetry and Remote Sensing, Volume 156〈/p〉 〈p〉Author(s): Wenxia Dai, Bisheng Yang, Xinlian Liang, Zhen Dong, Ronggang Huang, Yunsheng Wang, Wuyan Li〈/p〉 〈div xml:lang="en"〉 〈h5〉Abstract〈/h5〉 〈div〉〈p〉Airborne laser scanning (ALS) and terrestrial laser scanning (TLS) systems are effective ways to capture the 3D information of forests from complementary perspectives. Registration of the two sources of point clouds is necessary for various forestry applications. Since the forest point clouds show irregular and natural point distributions, standard registration methods working on geometric keypoints (e.g., points, lines, and planes) are likely to fail. Hence, we propose a novel method to register the ALS and TLS forest point clouds through density analysis of the crowns. The proposed method extracts mode-based keypoints by the mean shift method and aligns them by maximum likelihood estimation. Firstly, the differences in the point densities of the ALS and TLS crowns are minimized to produce analogous modes, which represent the local maxima of the underlying probability density function (PDF). The mode-based keypoints are then aligned through the coherent point drift (CPD) algorithm, which is independent of the descriptor similarities and considers the alignment as a maximum likelihood estimation problem. The sets of keypoints derived from the two data sources need not be equal. Finally, the recovered transformation is applied to the original point clouds and refined through the standard iterative closest point (ICP) algorithm. In contrast to some of the existing methods, the proposed method avoids the geometric description of the forest point clouds. Furthermore, additional information such as tree diameter or height is not required to evaluate the similarities. The experiments in this study were conducted in a Scandinavian boreal forest, located in Evo, Finland. The proposed method was tested on four datasets (ALS data: a circle with a diameter of 60 m, multi-scan TLS data: 32 × 32 m) with heterogeneous tree species and structures. The results showed that the proposed probabilistic-based method obtains a good performance with a 3D distance residual of 0.069 m, and improved the accuracy of the registration when compared with the existing methods.〈/p〉〈/div〉 〈/div〉

Description: 〈p〉Publication date: October 2019〈/p〉 〈p〉〈b〉Source:〈/b〉 ISPRS Journal of Photogrammetry and Remote Sensing, Volume 156〈/p〉 〈p〉Author(s): Mi Wang, Yufeng Cheng, Beibei Guo, Shuying Jin〈/p〉 〈div xml:lang="en"〉 〈h5〉Abstract〈/h5〉 〈div〉〈p〉The GaoFen6 (GF6) satellite is an optical remote sensing satellite in China’s GaoFen series that combines the imaging abilities of high resolution and wide field of view (WFV). The WFV camera, equipped on the GF6, can obtain an image 800 km in width and with a resolution of 16 m, which represents the highest level of observation swath width among similar satellites in the world. Guaranteeing the high geometric quality of the imagery obtained by the WFV camera for subsequent remote sensing applications is a prominent problem that must be solved. This paper presents the key geometry processing method for the WFV camera, in which the parameters relative determination method based on two selected reference bands and the sensor correction based on virtual collinear complementary metal oxide semiconductor (CMOS) with distortion are proposed. To overcome the challenge of matching the tie points between different bands with large radiation differences, two bands—rather than traditional one band—were chosen as the reference bands for the WFV camera to perform the parameters relative determination. The geometric influence of the virtual CMOS is fully analyzed in the along- and across-track directions. Then, the virtual CMOS with a real calibrated distortion was designed; based on this design, a sensor correction was applied to stitch the whole image generated by single images in each band. This experiment verified the effectiveness of the proposed parameters relative determination and sensor correction methods in improving the internal relative accuracy and the registration accuracy; moreover, the mosaicking accuracy is also guaranteed.〈/p〉〈/div〉 〈/div〉

Description: 〈p〉Publication date: October 2019〈/p〉 〈p〉〈b〉Source:〈/b〉 ISPRS Journal of Photogrammetry and Remote Sensing, Volume 156〈/p〉 〈p〉Author(s): Xiaoqiong Qin, Xiaoli Ding, Mingsheng Liao, Lu Zhang, Chisheng Wang〈/p〉 〈div xml:lang="en"〉 〈h5〉Abstract〈/h5〉 〈div〉〈p〉With the increasing operation time and environmental loads, a growing number of bridges are in poor conditions and suffer from continuous structural deformation. In order to understand how those deformations occurred and developed, the knowledge of bridge deformation characteristics and mechanisms is essential for users to manage and maintain bridges sustainably. The Differential Synthetic Aperture Radar Interferometry (DInSAR) technique is an effective tool to observe the bridge deformation and has achieved a few impressive results. Nevertheless, the previous studies often focused on bridges with simple structure and high coherence, avoiding the difficulties in identifying dense and accurate point-like targets (PTs) upon complexly structured bridges with more severe de-correlation effects. Moreover, the traditional two-dimensional (2D) deformation maps of complexly structured bridges are difficult for non-expert users to understand in multi-dimensional views, increasing the difficulty of deformation interpretation. Finally, simply analyzing all the PTs’ deformation velocities equivalently without considering the structural information is insufficient to explore the deformation characteristics and mechanisms of complexly structured bridges. Starting from these limitations, we proposed a bridge-tailored multi-temporal DInSAR approach, and for the first time, explored the deformation characteristics and mechanisms of the complexly structured arch bridge and cable-stayed bridge. The density and accuracy of detectable PTs on bridges are improved by implementing a structure coherence-driven PTs selection strategy. A simple and fast algorithm for 3D visualization of bridge results is accomplished through integrating orthorectified PTs’ measurements from multi-track SAR datasets. Based on the 3D products, the deformation characteristics of different types of bridges are investigated by identifying the damage sensitive points (DSPs) and revealing the global deformations, and the deformation mechanisms of different bridge components are explored by analyzing the different types of PTs classified based on the spatial-temporal semantics. It is derived that the deformation characteristics of bridges are associated with their structural and material characteristics, and the deformation mechanisms of their DSPs are similar. The subsidence of land surface and foundations control the deformation of the pier-DSPs which mainly reflect the trend deformation of bridges, whereas the temperature variation drives the deformation temporal evolution of the span-DSPs which represent the elastic deformation of bridges.〈/p〉〈/div〉 〈/div〉

Description: 〈p〉Publication date: October 2019〈/p〉 〈p〉〈b〉Source:〈/b〉 ISPRS Journal of Photogrammetry and Remote Sensing, Volume 156〈/p〉 〈p〉Author(s): Guilherme Silverio Aquino de Souza, Vicente Paulo Soares, Helio Garcia Leite, José Marinaldo Gleriani, Cibele Hummel do Amaral, Antônio Santana Ferraz, Marcus Vinicius de Freitas Silveira, 〈sup〉João〈/sup〉 Flávio Costa dos Santos, Sidney Geraldo Silveira Velloso, Getulio Fonseca Domingues, Simone Silva〈/p〉 〈div xml:lang="en"〉 〈h5〉Abstract〈/h5〉 〈div〉〈p〉Stem volume is a key attribute of 〈em〉Eucalyptus〈/em〉 forest plantations upon which decision-making is based at diverse levels of planning. Quantifying volume through remote sensing can support a proper management of forests. Because of limitations on spaceborne optical and synthetic aperture radar sensors, this study integrated both types of datasets assembled using support vector regression (SVR) to retrieve the stand volume of 〈em〉Eucalyptus〈/em〉 plantations. We assessed different combinations of sensors and a minimum number of plots to develop an SVR model. Finally, the best SVR performance was compared with other analytical methods already tested and in the literature: multilinear regression, artificial neural networks (ANN), and random forest (RF). Here, we introduce a test for comparative analysis of the performance of different methods. We found that SVR accurately predicted stem volume of Brazilian fast-growing 〈em〉Eucalyptus〈/em〉 forest plantations. Gaussian radial basis was the most suitable kernel function. Integrating the optical and L-band backscatter data increased the predictive accuracy compared to a single sensor model. Combining NIR-band data from ALOS AVNIR-2 and backscatter of L-band horizontal emitted and vertical received (HV) electric fields from ALOS PALSAR produced the most accurate SVR model (with an R〈sup〉2〈/sup〉 of 0.926 and root mean square error of 11.007 m〈sup〉3〈/sup〉/ha). The number of field plots sufficient for model development with non-redundant explanatory variables was 77. Under this condition, SVR performed similarly to ANN and outperformed the multiple linear regression and random forest methods.〈/p〉〈/div〉 〈/div〉

Description: 〈p〉Publication date: October 2019〈/p〉 〈p〉〈b〉Source:〈/b〉 ISPRS Journal of Photogrammetry and Remote Sensing, Volume 156〈/p〉 〈p〉Author(s): Hong Huang, Yule Duan, Haibo He, Guangyao Shi, Fulin Luo〈/p〉 〈div xml:lang="en"〉 〈h5〉Abstract〈/h5〉 〈div〉〈p〉Dimensionality reduction (DR) technology plays an important role in hyperspectral image (HSI) classification. However, many existing DR algorithms ignore the complex intrinsic structure in spatial domain and spectral domain of HSI. To address this issue, we put forward a spatial-spectral local discriminant projection (SSLDP) method based on the manifold learning theory and spatial consistency in HSI. In SSLDP, hyperspectral pixels are reconstructed by minimizing the weighted reconstruction errors to preserve the local geometric structure. Then, two weighted scatter matrices are designed to maintain the neighborhood structure in spatial domain and two reconstruction graphs are constructed to discover the local discriminant relationship in spectral domain. Finally, an objective function is designed for obtaining an optimal projection by compacting the spatial-spectral local intraclass points while separating the spatial-spectral local interclass points. The experiments performed on some real hyperspectral images, including the Indian Pines, PaviaU and Washington DC, demonstrate that the presented SSLDP algorithm is significantly superior to some state-of-the-art DR algorithms.〈/p〉〈/div〉 〈/div〉 〈div xml:lang="en"〉 〈h5〉Graphical abstract〈/h5〉 〈div〉〈p〉〈figure〉〈img src="https://ars.els-cdn.com/content/image/1-s2.0-S0924271619301595-ga1.jpg" width="462" alt="Graphical abstract for this article" title=""〉〈/figure〉〈/p〉〈/div〉 〈/div〉

Description: 〈p〉Publication date: December 2018〈/p〉 〈p〉〈b〉Source:〈/b〉 ISPRS Journal of Photogrammetry and Remote Sensing, Volume 146〈/p〉 〈p〉Author(s): Jean-François Tremblay, Martin Béland〈/p〉 〈div xml:lang="en"〉 〈h5〉Abstract〈/h5〉 〈div〉〈p〉Terrestrial laser scanning (TLS) often makes use of multiple scans in forests to allow for a complete view of a given area. Combining measurements from multiple locations requires accurate co-registration of the scans to a common reference coordinate system, which currently relies on markers, an often cumbersome process in forests. Existing algorithms for achieving marker-free registration of TLS scans in forests promise to significantly decrease field work time, but are not yet operational and their results have not been validated against traditional methods. Here we present a new implementation of an existing approach which runs in parallel mode and is able to process TLS data acquired over large forest areas. To validate our algorithm, point cloud registration matrices (translation and rotation) derived from our algorithm were compared to those obtained using reflective markers in multiple forest types. The results show that our approach can be used operationally in forests with relatively clear understory, and it provides accuracy similar to that obtained from using reflective markers. Furthermore, we identified factors that can lead to this approach falling short of providing acceptable results in terms of accuracy.〈/p〉〈/div〉 〈/div〉

Description: 〈p〉Publication date: December 2018〈/p〉 〈p〉〈b〉Source:〈/b〉 ISPRS Journal of Photogrammetry and Remote Sensing, Volume 146〈/p〉 〈p〉Author(s): Yumin Tan, Shuai Wang, Bo Xu, Jiabin Zhang〈/p〉 〈div xml:lang="en"〉 〈h5〉Abstract〈/h5〉 〈div〉〈p〉With the advent of unmanned aerial vehicle (UAV)-based photogrammetry and structure from motion (SFM) software, it is possible to obtain high-density point clouds of which the accuracy can meet the requirements of river bank monitoring. Ground filtering, i.e., removing the points belonging to above-ground objects, is an important process of digital terrain model (DTM) generation which is essential to river bank monitoring. Progressive morphological filter (PM) is a widely-adopted ground filtering algorithm and performs well with LiDAR data. However, it may incorrectly classify vegetation points as ground points when used to filter UAV-based photogrammetric point clouds because ground points beneath vegetation cannot be captured with the digital camera on-board UAV. In this study, we propose the improved progressive morphological filter (IPM) algorithm to improve the accuracy of ground filtering on UAV-based photogrammetric point clouds by introducing visible-band difference vegetation index (VDVI) to PM. The proposed IPM is subsequently evaluated along with the original PM algorithm and four other widely-used ground filtering algorithms in four test sites along the Yangtze River. The results show that IPM improves the overall accuracy from PM in all the four test sites, and produces the best results among the six ground filtering algorithms in three out of the four sites. IPM proves to be an effective ground filtering algorithm for UAV-based photogrammetric point clouds in river bank monitoring.〈/p〉〈/div〉 〈/div〉

Description: 〈p〉Publication date: December 2018〈/p〉 〈p〉〈b〉Source:〈/b〉 ISPRS Journal of Photogrammetry and Remote Sensing, Volume 146〈/p〉 〈p〉Author(s): Jennifer Roelens, Bernhard Höfle, Stefaan Dondeyne, Jos Van Orshoven, Jan Diels〈/p〉 〈div xml:lang="en"〉 〈h5〉Abstract〈/h5〉 〈div〉〈p〉Ditches are often absent in hydrographic geodatasets and their mapping would benefit from a cost and labor effective alternative to field surveys. We propose and evaluate an alternative that makes use of a high resolution LiDAR point cloud dataset. First the LiDAR points are classified as ditch and non-ditch points by means of a random forest classifier which considers subsets of the topographic and radiometric features provided by or derived from the LiDAR product. The LiDAR product includes for each georeferenced point, the elevation, the returned intensity value, and RGB values from simultaneously acquired aerial images. Next so-called ditch dropout points are reconstructed for the blind zones in the dataset using a new geometric approach. Finally, LiDAR ditch points and dropouts are assembled into ditch objects (2D-polygons and their derived centre lines). The procedure was evaluated for a grassland and a peri-urban agricultural area in Flanders, Belgium. A good point classification was obtained (Kappa = 0.77 for grassland and 0.73 for peri-urban area) by using all the features derived from the LiDAR product, whereby the geometric features had the greatest influence. However, even better results were obtained when the radiometric component of the LiDAR product was also taken into account. For the tested models for the extraction of ditch centre lines, the best resulted in an error of omission of 0.03 and an error of commission of 0.08 for the grassland study area and an error of omission of 0.14 and an error of commission of 0.07 for the peri-urban study area.〈/p〉〈/div〉 〈/div〉

Description: 〈p〉Publication date: May 2019〈/p〉 〈p〉〈b〉Source:〈/b〉 ISPRS Journal of Photogrammetry and Remote Sensing, Volume 151〈/p〉 〈p〉Author(s): Fariba Mohammadimanesh, Bahram Salehi, Masoud Mahdianpari, Eric Gill, Matthieu Molinier〈/p〉 〈div xml:lang="en"〉 〈h5〉Abstract〈/h5〉 〈div〉〈p〉Despite the application of state-of-the-art fully Convolutional Neural Networks (CNNs) for semantic segmentation of very high-resolution optical imagery, their capacity has not yet been thoroughly examined for the classification of Synthetic Aperture Radar (SAR) images. The presence of speckle noise, the absence of efficient feature expression, and the limited availability of labelled SAR samples have hindered the application of the state-of-the-art CNNs for the classification of SAR imagery. This is of great concern for mapping complex land cover ecosystems, such as wetlands, where backscattering/spectrally similar signatures of land cover units further complicate the matter. Accordingly, we propose a new Fully Convolutional Network (FCN) architecture that can be trained in an end-to-end scheme and is specifically designed for the classification of wetland complexes using polarimetric SAR (PolSAR) imagery. The proposed architecture follows an encoder-decoder paradigm, wherein the input data are fed into a stack of convolutional filters (encoder) to extract high-level abstract features and a stack of transposed convolutional filters (decoder) to gradually up-sample the low resolution output to the spatial resolution of the original input image. The proposed network also benefits from recent advances in CNN designs, namely the addition of inception modules and skip connections with residual units. The former component improves multi-scale inference and enriches contextual information, while the latter contributes to the recovery of more detailed information and simplifies optimization. Moreover, an in-depth investigation of the learned features via opening the 〈em〉black box〈/em〉 demonstrates that convolutional filters extract discriminative polarimetric features, thus mitigating the limitation of the feature engineering design in PolSAR image processing. Experimental results from full polarimetric RADARSAT-2 imagery illustrate that the proposed network outperforms the conventional random forest classifier and the state-of-the-art FCNs, such as FCN-32s, FCN-16s, FCN-8s, and SegNet, both visually and numerically for wetland mapping.〈/p〉〈/div〉 〈/div〉 〈div xml:lang="en"〉 〈h5〉Graphical abstract〈/h5〉 〈div〉〈p〉〈figure〉〈img src="https://ars.els-cdn.com/content/image/1-s2.0-S092427161930084X-ga1.jpg" width="500" alt="Graphical abstract for this article" title=""〉〈/figure〉〈/p〉〈/div〉 〈/div〉

Description: 〈p〉Publication date: May 2019〈/p〉 〈p〉〈b〉Source:〈/b〉 ISPRS Journal of Photogrammetry and Remote Sensing, Volume 151〈/p〉 〈p〉Author(s): Wenzhi Zhao, Yanchen Bo, Jiage Chen, Dirk Tiede, Blaschke Thomas, William J. Emery〈/p〉 〈div xml:lang="en"〉 〈h5〉Abstract〈/h5〉 〈div〉〈p〉Urban scenes refer to city blocks which are basic units of megacities, they play an important role in citizens’ welfare and city management. Remote sensing imagery with largescale coverage and accurate target descriptions, has been regarded as an ideal solution for monitoring the urban environment. However, due to the heterogeneity of remote sensing images, it is difficult to access their geographical content at the object level, let alone understanding urban scenes at the block level. Recently, deep learning-based strategies have been applied to interpret urban scenes with remarkable accuracies. However, the deep neural networks require a substantial number of training samples which are hard to satisfy, especially for high-resolution images. Meanwhile, the crowed-sourced Open Street Map (OSM) data provides rich annotation information about the urban targets but may encounter the problem of insufficient sampling (limited by the places where people can go). As a result, the combination of OSM and remote sensing images for efficient urban scene recognition is urgently needed. In this paper, we present a novel strategy to transfer existing OSM data to high-resolution images for semantic element determination and urban scene understanding. To be specific, the object-based convolutional neural network (OCNN) can be utilized for geographical object detection by feeding it rich semantic elements derived from OSM data. Then, geographical objects are further delineated into their functional labels by integrating points of interest (POIs), which contain rich semantic terms, such as commercial or educational labels. Lastly, the categories of urban scenes are easily acquired from the semantic objects inside. Experimental results indicate that the proposed method has an ability to classify complex urban scenes. The classification accuracies of the Beijing dataset are as high as 91% at the object-level and 88% at the scene level. Additionally, we are probably the first to investigate the object level semantic mapping by incorporating high-resolution images and OSM data of urban areas. Consequently, the method presented is effective in delineating urban scenes that could further boost urban environment monitoring and planning with high-resolution images.〈/p〉〈/div〉 〈/div〉

Description: 〈p〉Publication date: May 2019〈/p〉 〈p〉〈b〉Source:〈/b〉 ISPRS Journal of Photogrammetry and Remote Sensing, Volume 151〈/p〉 〈p〉Author(s): Zihan Liu, Wenfeng Zhan, Jiameng Lai, Falu Hong, Jinling Quan, Benjamin Bechtel, Fan Huang, Zhaoxu Zou〈/p〉 〈div xml:lang="en"〉 〈h5〉Abstract〈/h5〉 〈div〉〈p〉Annual temperature cycle (ATC) models enable the multi-timescale analysis of land surface temperature (LST) dynamics and are therefore valuable for various applications. However, the currently available ATC models focus either on prediction accuracy or on generalization ability and a flexible ATC modelling framework for different numbers of thermal observations is lacking. Here, we propose a hybrid ATC model (ATCH) that considers both prediction accuracy and generalization ability; our approach combines multiple harmonics with a linear function of LST-related factors, including surface air temperature (SAT), NDVI, albedo, soil moisture, and relative humidity. Based on the proposed ATCH, various parameter-reduction approaches (PRAs) are designed to provide model derivatives which can be adapted to different scenarios. Using Terra/MODIS daily LST products as evaluation data, the ATCH is compared with the original sinusoidal ATC model (termed the ATCO) and its variants, and with two frequently-used gap-filling methods (Regression Kriging Interpolation (RKI) and the Remotely Sensed DAily land Surface Temperature reconstruction (RSDAST)), under clear-sky conditions. In addition, under overcast conditions, the LSTs generated by ATCH are directly compared with 〈em〉in-situ〈/em〉 LST measurements. The comparisons demonstrate that the ATCH increases the prediction accuracy and the overall RMSE is reduced by 1.8 and 0.7 K when compared with the ATCO during daytime and nighttime, respectively. Moreover, the ATCH shows better generalization ability than the RKI and behaves better than the RSDAST when the LST gap size is spatially large and/or temporally long. By employing LST-related controls (e.g., the SAT and relative humidity) under overcast conditions, the ATCH can better predict the LSTs under clouds than approaches that only adopt clear-sky information as model inputs. Further attribution analysis implies that incorporating a sinusoidal function (ASF), the SAT, NDVI, and other LST-related factors, provides respective contributions of around 16%, 40%, 15%, and 30% to the improved accuracy. Our analysis is potentially useful for designing PRAs for various practical needs, by reducing the smallest contribution factor each time. We conclude that the ATCH is valuable for further improving the quality of LST products and can potentially enhance the time series analysis of land surfaces and other applications.〈/p〉〈/div〉 〈/div〉

Description: 〈p〉Publication date: May 2019〈/p〉 〈p〉〈b〉Source:〈/b〉 ISPRS Journal of Photogrammetry and Remote Sensing, Volume 151〈/p〉 〈p〉Author(s): Yanfei Zhong, Yao Xu, Xinyu Wang, Tianyi Jia, Guisong Xia, Ailong Ma, Liangpei Zhang〈/p〉 〈div xml:lang="en"〉 〈h5〉Abstract〈/h5〉 〈div〉〈p〉In mid- and high-latitude regions, district heating systems (DHSs) are major heat supply solutions to both local industry and citizens. Pipeline leakage detection is therefore important for monitoring the condition of DHSs and promoting energy efficiency. In this paper, a saliency analysis method is presented for DHS pipeline leakage detection using remotely sensed infrared imagery, visible imagery, and geographic information system (GIS) data. In the saliency-based DHS leakage detection method, the infrared saliency map is created to enhance the leakage targets, and the pipeline location information extracted from the GIS data or the visible imagery acts as a distribution prior to reject false alarms. Finally, adaptive target segmentation by maximum entropy permits the automatic detection of potential leakage targets in the final fused saliency map. The approach was validated on three data sets acquired in Gävle in Sweden and Datong in China, with the heating leakages indicated by human analysts and field validation. The leakage detection accuracy of the new approach with a reduced false alarm rate is better than the previous methods. The results suggest that the proposed approach for DHS leakage detection from remotely sensed thermal infrared data has great potential for monitoring DHS conditions in mid- and high-latitude regions.〈/p〉〈/div〉 〈/div〉

Description: 〈p〉Publication date: January 2019〈/p〉 〈p〉〈b〉Source:〈/b〉 ISPRS Journal of Photogrammetry and Remote Sensing, Volume 147〈/p〉 〈p〉Author(s): Danfeng Hong, Naoto Yokoya, Nan Ge, Jocelyn Chanussot, Xiao Xiang Zhu〈/p〉 〈div xml:lang="en"〉 〈h5〉Abstract〈/h5〉 〈div〉〈p〉In this paper, we aim at tackling a general but interesting cross-modality feature learning question in remote sensing community—〈em〉can a limited amount of highly-discriminative (e.g., hyperspectral) training data improve the performance of a classification task using a large amount of poorly-discriminative (e.g., multispectral) data?〈/em〉 Traditional semi-supervised manifold alignment methods do not perform sufficiently well for such problems, since the hyperspectral data is very expensive to be largely collected in a trade-off between time and efficiency, compared to the multispectral data. To this end, we propose a novel semi-supervised cross-modality learning framework, called learnable manifold alignment (LeMA). LeMA learns a joint graph structure directly from the data instead of using a given fixed graph defined by a Gaussian kernel function. With the learned graph, we can further capture the data distribution by graph-based label propagation, which enables finding a more accurate decision boundary. Additionally, an optimization strategy based on the alternating direction method of multipliers (ADMM) is designed to solve the proposed model. Extensive experiments on two hyperspectral-multispectral datasets demonstrate the superiority and effectiveness of the proposed method in comparison with several state-of-the-art methods.〈/p〉〈/div〉 〈/div〉

Description: 〈p〉Publication date: January 2019〈/p〉 〈p〉〈b〉Source:〈/b〉 ISPRS Journal of Photogrammetry and Remote Sensing, Volume 147〈/p〉 〈p〉Author(s): Zhipeng Cai, Tat-Jun Chin, Alvaro Parra Bustos, Konrad Schindler〈/p〉 〈div xml:lang="en"〉 〈h5〉Abstract〈/h5〉 〈div〉 〈p〉Point cloud registration is a fundamental problem in 3D scanning. In this paper, we address the frequent special case of registering terrestrial LiDAR scans (or, more generally, levelled point clouds). Many current solutions still rely on the Iterative Closest Point (ICP) method or other heuristic procedures, which require good initializations to succeed and/or provide no guarantees of success. On the other hand, exact or optimal registration algorithms can compute the best possible solution without requiring initializations; however, they are currently too slow to be practical in realistic applications.〈/p〉 〈p〉Existing optimal approaches ignore the fact that in routine use the relative rotations between scans are constrained to the azimuth, via the built-in level compensation in LiDAR scanners. We propose a novel, optimal and computationally efficient registration method for this 4DOF scenario. Our approach operates on candidate 3D keypoint correspondences, and contains two main steps: (1) a deterministic selection scheme that significantly reduces the candidate correspondence set in a way that is guaranteed to preserve the optimal solution; and (2) a fast branch-and-bound (BnB) algorithm with a novel polynomial-time subroutine for 1D rotation search, that quickly finds the optimal alignment for the reduced set. We demonstrate the practicality of our method on realistic point clouds from multiple LiDAR surveys.〈/p〉 〈/div〉 〈/div〉

Description: 〈p〉Publication date: February 2019〈/p〉 〈p〉〈b〉Source:〈/b〉 ISPRS Journal of Photogrammetry and Remote Sensing, Volume 148〈/p〉 〈p〉Author(s): Leiqiu Hu, Jochen Wendel〈/p〉 〈div xml:lang="en"〉 〈h5〉Abstract〈/h5〉 〈div〉〈p〉Remote thermal radiative observations over metropolitan areas are subject to an angular-dependent variation, known as the directional thermal anisotropy. The 3D urban surface morphology is one key factor in determining the magnitude and temporal variation of thermal anisotropy. This study uses 3D building data and the Town Energy Balance model (TEB) to explore the impact of morphological variability on diurnal anisotropy patterns, and quantifies errors introduced by a simplification of urban morphology as an array of evenly distributed uniform cubes. Results from the comparison of two representative urban districts in Brooklyn and midtown Manhattan, New York City reveal distinct diurnal anisotropy patterns. Daytime anisotropy varies more time-sensitively over the compact high-rise district of Manhattan, although the maximum effective anisotropy (the maximum contrast of directional anisotropy) is smaller than Brooklyn, which is related to a reduced contrast among wall temperatures. A stronger angular effect at night is found as the aspect-ratio increases. The anisotropy is further simulated at the Moderate Resolution Imaging Spectroradiometer (MODIS) overpass time and sensor-surface relative geometry over these two morphologic samples. The sensitivity test unravels that the effective anisotropy monotonically increases with a greater aspect ratio for MODIS nighttime overpasses, while the daytime pattern is more complex with a single- or double-peak distribution depending on the solar angle (or time of day). Finally, the variation of building height and size is important in determining the anisotropy from comparing simulations of a realistic 3D building model and a simplified urban morphology as cube array. The morphological simplification can lead to a higher discrepancy in these cases with a high aspect ratio or small sky view factor for both daytime and nighttime. The proposed 3D-computer-graphics approach is computationally affordable for the seen surface estimation and can be applied to IFOVs across a relatively large urban area. Its flexibility in integrating various levels of 3D urban surface complexity makes it a promising tool for correcting the urban thermal anisotropy from satellite observations in the future.〈/p〉〈/div〉 〈/div〉 〈div xml:lang="en"〉 〈h5〉Graphical abstract〈/h5〉 〈div〉〈p〉〈figure〉〈img src="https://ars.els-cdn.com/content/image/1-s2.0-S0924271618303332-ga1.jpg" width="499" alt="Graphical abstract for this article" title=""〉〈/figure〉〈/p〉〈/div〉 〈/div〉

Description: 〈p〉Publication date: January 2019〈/p〉 〈p〉〈b〉Source:〈/b〉 ISPRS Journal of Photogrammetry and Remote Sensing, Volume 147〈/p〉 〈p〉Author(s): 〈/p〉

Description: 〈p〉Publication date: January 2019〈/p〉 〈p〉〈b〉Source:〈/b〉 ISPRS Journal of Photogrammetry and Remote Sensing, Volume 147〈/p〉 〈p〉Author(s): Andreise Moreira, Denise Cybis Fontana, Tatiana Mora Kuplich〈/p〉 〈div xml:lang="en"〉 〈h5〉Abstract〈/h5〉 〈div〉〈p〉The aim of this study was to describe the phenology of different types of grasslands from the Pampa and Mata Atlântica biomes in the southern region of Brazil and its relation with meteorological variables, in a time series of EVI/MODIS data using the Wavelet approach (Transform – WT - and Coherence - WC). There is a lack of studies focusing on how climate variability influences the phenology of different Brazilian Pampa grassland typologies. This information is essential to describe the spatio-temporal dynamics of grasslands and contribute to actions of sustainable management and conservation strategies, threatened by crops, forestry, and expansion of overgrazed cattle farms. A series of EVI/MODIS acquired from February 2000 to December 2014, totaling 342 images, were sampled for each of the 10 grassland typologies at the study area. Mean EVI data and the WT indicated when and where changes in the grassland phenological dynamics occurred. The WC, applied to the EVI/MODIS time series with (i) rainfall and (ii) air temperature, helped to identify the correlation between the data. Two well - defined cycles were identified: annual, ranging from 1 to 23 observations, and interannual, from 92 to 184 observations. The different grassland typologies showed similar phenological patterns, although some spatial dependency was observed and related to the different soil and terrain morphometry at the study area. The influence of those abiotic factors on the grassland vegetation, phenological events and their expression on the EVI was also spatially dependent and strongly linked to weather conditions (e.g., the permanence of humidity after rainfall in shallow soils) and climate. The correlation between EVI and air temperature was stronger in the annual cycle for all grassland typologies. For the interannual cycles, El Niño and La Niña events caused higher correlation between EVI data and rainfall.〈/p〉〈/div〉 〈/div〉

Description: 〈p〉Publication date: January 2019〈/p〉 〈p〉〈b〉Source:〈/b〉 ISPRS Journal of Photogrammetry and Remote Sensing, Volume 147〈/p〉 〈p〉Author(s): Fabio Castaldi, Andreas Hueni, Sabine Chabrillat, Kathrin Ward, Gabriele Buttafuoco, Bart Bomans, Kristin Vreys, Maximilian Brell, Bas van Wesemael〈/p〉 〈div xml:lang="en"〉 〈h5〉Abstract〈/h5〉 〈div〉〈p〉The short revisit time of the Sentinel-2 (S2) constellation entails a large availability of remote sensing data, but S2 data have been rarely used to predict soil organic carbon (SOC) content. Thus, this study aims at comparing the capability of multispectral S2 and airborne hyperspectral remote sensing data for SOC prediction, and at the same time, we investigated the importance of spectral and spatial resolution through the signal-to-noise ratio (SNR), the variable importance in the prediction (VIP) models and the spatial variability of the SOC maps at field and regional scales. We tested the capability of the S2 data to predict SOC in croplands with quite different soil types and parent materials in Germany, Luxembourg and Belgium, using multivariate statistics and local ground calibration with soil samples. We split the calibration dataset into sub-regions according to soil maps and built a multivariate regression model within each sub-region. The prediction accuracy obtained by S2 data is generally slightly lower than that retrieved by airborne hyperspectral data. The ratio of performance to deviation (RPD) is higher than 2 in Luxembourg (2.6) and German (2.2) site, while it is 1.1 in the Belgian area. After the spectral resampling of the airborne data according to S2 band, the prediction accuracy did not change for four out of five of the sub-regions. The variable importance values obtained by S2 data showed the same trend as the airborne VIP values, while the importance of SWIR bands decreased using airborne data resampled according the S2 bands. These differences of VIP values can be explained by the loss of spectral resolution as compared to APEX data and the strong difference in terms of SNR between the SWIR region and other spectral regions. The investigation on the spatial variability of the SOC maps derived by S2 data has shown that the spatial resolution of S2 is adequate to describe SOC variability both within field and at regional scale.〈/p〉〈/div〉 〈/div〉

Description: 〈p〉Publication date: December 2018〈/p〉 〈p〉〈b〉Source:〈/b〉 ISPRS Journal of Photogrammetry and Remote Sensing, Volume 146〈/p〉 〈p〉Author(s): Liang Cheng, Hao Xu, Shuyi Li, Yanming Chen, Fangli Zhang, Manchun Li〈/p〉

Description: 〈p〉Publication date: February 2019〈/p〉 〈p〉〈b〉Source:〈/b〉 ISPRS Journal of Photogrammetry and Remote Sensing, Volume 148〈/p〉 〈p〉Author(s): Lin Cao, Nicholas C. Coops, Yuan Sun, Honghua Ruan, Guibin Wang, Jinsong Dai, Guanghui She〈/p〉 〈div xml:lang="en"〉 〈h5〉Abstract〈/h5〉 〈div〉〈p〉The Bamboo species accounts for almost 1% of the Earth’s forested area with an exceptionally fast growth peaking up to 7.5–100 cm per day during the growing period, making it an unique species with respect to measuring and monitoring using conventional forest inventory tools. In addition their widespread coverage and quick growth make them a critical component of the terrestrial carbon cycle and for mitigating the impacts of climate change. In this study, the capability of using airborne Light Detection and Ranging (LiDAR) data for estimating canopy structure and biomass of Moso bamboo (〈em〉Phyllostachys pubescens〈/em〉) was assessed, which is one of the most valuable and widely distributed bamboo species in the subtropical forests of south China. To do so, we first evaluated the accuracy of using LiDAR data to interpolate the underlying ground terrain under bamboo forests and developed uncertainty surfaces using both LiDAR-derived vegetation and topographic metrics and a Random Forest (RF) classifier. Second, we utilized Principal Component Analysis (PCA) to quantify the variation of the vertical distribution of LiDAR-derived effective Leaf Area Index (LAI) of bamboo stands, and fitted regression models between selected LiDAR metrics and the field-measured attributes such mean height, DBH and biomass components (i.e., culm, branch, foliage and aboveground biomass (AGB)) across a range of management strategies. Once models were developed, the results were spatially extrapolated and compared across the bamboo stands. Results indicated that the LiDAR interpolated DTMs were accurate even under the dense intensively managed bamboo stands (RMSE = 0.117–0.126 m) as well as under secondary stands (RMSE = 0.102 m) with rugged terrain and near-ground dense vegetation. The development of uncertainty maps of terrain was valuable when examining the magnitude and spatial distribution of potential errors in the DTMs. The middle height intervals (i.e., HI4 and HI5) within the bamboo cumulative effective LAI profiles explained more variances by PCA analysis in the bamboo stands. Moso bamboo AGB was well predicted by the LiDAR metrics (R〈sup〉2〈/sup〉 = 0.59–0.87, rRMSE = 11.92–21.11%) with percentile heights (〈em〉h〈/em〉〈sub〉25〈/sub〉-〈em〉h〈/em〉〈sub〉95〈/sub〉) and the coefficient of variation of height (〈em〉h〈/em〉〈sub〉cv〈/sub〉) having the highest relative importances for estimating AGB and culm biomass. The 〈em〉h〈/em〉〈sub〉cv〈/sub〉 explained the most variance in branch and foliage biomass. According to the spatial extrapolation results, areas of relatively low biomass were found on secondary stands (AGB = 49.42 ± 14.16 Mg ha〈sup〉−1〈/sup〉), whereas the intensively managed stands (AGB = 173.47 ± 34.16 Mg ha〈sup〉−1〈/sup〉) have much higher AGB and biomass components, followed by the extensively managed bamboo stands (AGB = 67.61 ± 13.10 Mg ha〈sup〉−1〈/sup〉). This study demonstrated the potential benefits of using airborne LiDAR to accurately derive high resolution DTMs, characterize vertical structure of canopy and estimate the magnitude and distribution of biomass within Moso bamboo forests, providing key data for regional ecological, environmental and global carbon cycle models.〈/p〉〈/div〉 〈/div〉

Description: 〈p〉Publication date: February 2019〈/p〉 〈p〉〈b〉Source:〈/b〉 ISPRS Journal of Photogrammetry and Remote Sensing, Volume 148〈/p〉 〈p〉Author(s): Li Li, Jingmin Tu, Ye Gong, Jian Yao, Jie Li〈/p〉 〈div xml:lang="en"〉 〈h5〉Abstract〈/h5〉 〈div〉〈p〉For multiple orthoimages mosaicking, the detection of an optimal seamline in an overlapped region and the generation of a seamline network are two key issues for creating a seamless and pleasant large-scale digital orthophoto map. In this paper, a novel system is proposed to generate the large-scale orthophoto by mosaicking multiple orthoimages via Graph cuts. The proposed system is comprised of two parts. In the first part, to ensure that the detected seamline avoids crossing the obvious objects, a novel foreground segmentation-based approach is proposed to detect the optimal seamline for two adjacent images. The foreground objects are segmented from the overlapped region at the superpixel level followed by the pixel-level seamline optimization. In the second part, we propose a novel seamline network generation approach to produce the large-scale orthophoto by mosaicking multiple orthoimages. The pairwise and junction regions extracted from the initial network are refined using two-label and multi-label Graph cuts, respectively. The key advantage of our proposed seamline network is that junction points can be automatically and optimally found using the multi-label Graph cuts. The experimental results on two groups of orthoimages show that our proposed system can generate high-quality seamline networks with less artifacts, and that it outperforms the state-of-the-art algorithm and the commercial software based on visual comparison and statistical evaluation.〈/p〉〈/div〉 〈/div〉

Description: 〈p〉Publication date: February 2019〈/p〉 〈p〉〈b〉Source:〈/b〉 ISPRS Journal of Photogrammetry and Remote Sensing, Volume 148〈/p〉 〈p〉Author(s): Michael Schmitt, Gerald Baier, Xiao Xiang Zhu〈/p〉 〈div xml:lang="en"〉 〈h5〉Abstract〈/h5〉 〈div〉〈p〉This article investigates the potential of nonlocally filtered pursuit monostatic TanDEM-X data for coastline detection in comparison to conventional TanDEM-X data, i.e. image pairs acquired in repeat-pass or bistatic mode. For this task, an unsupervised coastline detection procedure based on scale-space representations and 〈em〉K〈/em〉-medians clustering as well as morphological image post-processing is proposed. Since this procedure exploits a clear discriminability of “dark” and “bright” appearances of water and land surfaces, respectively, in both SAR amplitude and coherence imagery, TanDEM-X InSAR data acquired in pursuit monostatic mode is expected to provide a promising benefit. In addition, we investigate the benefit introduced by a utilization of a non-local InSAR filter for amplitude denoising and coherence estimation instead of a conventional box-car filter. Experiments carried out on real TanDEM-X pursuit monostatic data confirm our expectations and illustrate the advantage of the employed data configuration over conventional TanDEM-X products for automatic coastline detection.〈/p〉〈/div〉 〈/div〉

Description: 〈p〉Publication date: February 2019〈/p〉 〈p〉〈b〉Source:〈/b〉 ISPRS Journal of Photogrammetry and Remote Sensing, Volume 148〈/p〉 〈p〉Author(s): Chisheng Wang, Qiqi Shu, Xinyu Wang, Bo Guo, Peng Liu, Qingquan Li〈/p〉 〈div xml:lang="en"〉 〈h5〉Abstract〈/h5〉 〈div〉〈p〉The outstanding accuracy and spatial resolution of airborne light detection and ranging (LiDAR) systems allow for very detailed urban monitoring. Classification is a crucial step in LiDAR data processing, as many applications, e.g., 3D city modeling, building extraction, and digital elevation model (DEM) generation, rely on classified results. In this study, we present a novel LiDAR classification approach that uses simple pixel comparison features instead of the manually designed features used in many previous studies. The proposed features are generated by the computed height difference between two randomly selected neighboring pixels. In this way, the feature design does not require prior knowledge or human effort. More importantly, the features encode contextual information and are extremely quick to compute. We apply a random forest classifier to these features and a majority analysis postprocessing step to refine the classification results. The experiments undertaken in this study achieved an overall accuracy of 87.2%, which can be considered good given that only height information from the LiDAR data was used. The results were better than those obtained by replacing the proposed features with five widely accepted man-made features. We conducted algorithm parameter setting tests and an importance analysis to explore how the algorithm works. We found that the pixel pairs directing along the object structure and with a distance of the approximate object size can generate more discriminative pixel comparison features. Comparison with other benchmark results shows that this algorithm can approach the performance of state-of-the-art deep learning algorithms and exceed them in computational efficiency. We conclude that the proposed algorithm has high potential for urban LiDAR classification.〈/p〉〈/div〉 〈/div〉

Description: 〈p〉Publication date: February 2019〈/p〉 〈p〉〈b〉Source:〈/b〉 ISPRS Journal of Photogrammetry and Remote Sensing, Volume 148〈/p〉 〈p〉Author(s): Fan Xue, Weisheng Lu, Christopher J. Webster, Ke Chen〈/p〉 〈div xml:lang="en"〉 〈h5〉Abstract〈/h5〉 〈div〉〈p〉Symmetry is ubiquitous in architecture, across both time and place. Automated architectural symmetry detection (ASD) from a data source is not only an intriguing inquiry in its own right, but also a step towards creation of semantically rich building and city information models with applications in architectural design, construction management, heritage conservation, and smart city development. While recent advances in sensing technologies provide inexpensive yet high-quality architectural 3D point clouds, existing methods of ASD from these data sources suffer several weaknesses including noise sensitivity, inaccuracy, and high computational loads. This paper aims to develop a novel derivative-free optimization (DFO)-based approach for effective ASD. It does so by firstly transforming ASD into a nonlinear optimization problem involving architectural regularity and topology. An in-house ODAS (Optimization-based Detection of Architectural Symmetries) approach is then developed to solve the formulated problem using a set of state-of-the-art DFO algorithms. Efficiency, accuracy, and robustness of ODAS are gauged from the experimental results on nine sets of real-life architectural 3D point clouds, with the computational time for ASD from 1.4 million points only 3.7 s and increasing in a sheer logarithmic order against the number of points. The contributions of this paper are threefold. Firstly, formulating ASD as a nonlinear optimization problem constitutes a methodological innovation. Secondly, the provision of up-to-date, open source DFO algorithms allows benchmarking in the future development of free, fast, accurate, and robust approaches for ASD. Thirdly, the ODAS approach can be directly used to develop building and city information models for various value-added applications.〈/p〉〈/div〉 〈/div〉

Description: 〈p〉Publication date: February 2019〈/p〉 〈p〉〈b〉Source:〈/b〉 ISPRS Journal of Photogrammetry and Remote Sensing, Volume 148〈/p〉 〈p〉Author(s): J. Koskinen, U. Leinonen, A. Vollrath, A. Ortmann, E. Lindquist, R. d'Annunzio, A. Pekkarinen, N. Käyhkö〈/p〉 〈div xml:lang="en"〉 〈h5〉Abstract〈/h5〉 〈div〉〈p〉Recent years have witnessed the practical value of open-access Earth observation data catalogues and software in land and forest mapping. Combined with cloud-based computing resources, and data collection through the crowd, these solutions have substantially improved possibilities for monitoring changes in land resources, especially in areas with difficult accessibility and data scarcity. In this study, we developed and tested a participatory mapping methodology utilizing the open data catalogues and cloud computing capacity to map the previously unknown extent and species composition of forest plantations in the Southern Highlands area of Tanzania, a region experiencing a rapid growth of smallholder-owned woodlots. A large set of reference data, focusing on forest plantation coverage, species and age information distribution, was collected in a two-week participatory GIS campaign where 22 Tanzanian experts interpreted very high-resolution satellite images in Google Earth with the Open Foris Collect Earth tool developed by the Food and Agriculture Organization of the United Nations. The collected samples were used as training data to classify a multi-sensor image stack of Landsat 8 (2013–2015), Sentinel-2 (2015–2016), Sentinel-1 (2015), and SRTM derived elevation and slope data layers into a 30 m resolution forest plantation map in Google Earth Engine. The results show that the forest plantation area was estimated with high overall accuracy (85%). The interpretation accuracy of local experts was high considering general definition of forest plantation declining with increased details in interpretation attributes. The results showcase the unique value of local expert participation, enabling the collection of thousands of reference samples over a large geographical area in a short period of time simultaneously building the capacity of the experts. However, sufficient training prior to the data collection is crucial for the interpretation success especially when detailed interpretation is conducted in complex landscapes. Since the methodology is built on open-access data and software, it presents a highly feasible solution for repetitive land resource mapping applicable at different spatial scales globally.〈/p〉〈/div〉 〈/div〉

Description: 〈p〉Publication date: February 2019〈/p〉 〈p〉〈b〉Source:〈/b〉 ISPRS Journal of Photogrammetry and Remote Sensing, Volume 148〈/p〉 〈p〉Author(s): Xuebo Yang, Cheng Wang, Feifei Pan, Sheng Nie, Xiaohuan Xi, Shezhou Luo〈/p〉 〈div xml:lang="en"〉 〈h5〉Abstract〈/h5〉 〈div〉〈p〉Leaf area index (LAI) is an important vegetation structure parameter in terrestrial ecosystem modeling. Although the spaceborne Geoscience Laser Altimeter System (GLAS) on board the Ice, Cloud and land Elevation Satellite (ICESat) has been proved to have potential for deriving forest LAI, previous methods were only applicable to estimate the effective LAI. In this study, a physical method based on the gap fraction model was proposed to retrieve the LAI correcting the between-crown clumping in discontinuous forest using GLAS full-waveform data. Landsat TM imagery was utilized as auxiliary data for providing crown cover information within the footprint. Using the gap probability from GLAS data and the crown coverage fraction from Landsat imagery, the method corrects the between-crown clumping, which has been proved to contribute most to the total clumping effect, and accurately estimates the LAI in discontinuous forest (〈em〉R〈/em〉〈sup〉2〈/sup〉 = 0.83, RMSE = 0.39, 〈em〉n〈/em〉 = 47). Additionally, the forest LAI underestimation caused by between-crown clumping was analyzed in practice and theory. Results show that the between-crown clumping has a nonnegligible influence on forest LAI estimation in cases that the forest area within the footprint is close to the nonforest area, and the average LAI of individual tree is high. This study may shed some light on the development of clumping effect and quantitative LAI inversion models.〈/p〉〈/div〉 〈/div〉

Description: 〈p〉Publication date: January 2019〈/p〉 〈p〉〈b〉Source:〈/b〉 ISPRS Journal of Photogrammetry and Remote Sensing, Volume 147〈/p〉 〈p〉Author(s): Tengfei Su〈/p〉 〈div xml:lang="en"〉 〈h5〉Abstract〈/h5〉 〈div〉〈p〉In high resolution remote sensing imagery (HRI), the sizes of different geo-objects often vary greatly, posing serious difficulties to their successful segmentation. Although existent segmentation approaches have provided some solutions to this problem, the complexity of HRI may still lead to great challenges for previous methods. In order to further enhance the quality of HRI segmentation, this paper proposes a new segmentation algorithm based on scale-variable region merging. Scale-variable means that the scale parameters (SP) adopted for segmentation are adaptively estimated, so that geo-objects of various sizes can be better segmented out. To implement the proposed technique, 3 steps are designed. The first step produces a coarse-segmentation result with slight degree of under segmentation error. This is achieved by segmenting a half size image with the global optimal SP. Such a SP is determined by using the image of original size. In the second step, structural and spatial contextual information is extracted from the coarse-segmentation, enabling the estimation of variable SPs. In the last step, a region merging process is initiated, and the SPs used to terminate this process are estimated based on the information obtained in the second step. The proposed method was tested by using 3 scenes of HRI with different landscape patterns. Experimental results indicated that our approach produced good segmentation accuracy, outperforming some competitive methods in comparison.〈/p〉〈/div〉 〈/div〉

Description: 〈p〉Publication date: February 2019〈/p〉 〈p〉〈b〉Source:〈/b〉 ISPRS Journal of Photogrammetry and Remote Sensing, Volume 148〈/p〉 〈p〉Author(s): Nan Li, Chun Liu, Norbert Pfeifer〈/p〉 〈div xml:lang="en"〉 〈h5〉Abstract〈/h5〉 〈div〉〈p〉We propose a contextual label-smoothing method to improve the LiDAR classification accuracy in a post-processing step. Under the framework of global graph-structured regularization, we enhance the effectiveness of label smoothing from two aspects. First, each point can collect sufficient label-relevant neighborhood information to verify its label based on an optimal graph. Second, the input label probability set is improved by probabilistic label relaxation to be more consistent with the spatial context. With this optimal graph and reliable label probability set, the final labels are computed by graph-structured regularization. We demonstrate the contextual label-smoothing approach on two separate urban airborne LiDAR datasets with complex urban scenes. Significant improvements in the classification accuracies are achieved without losing small objects (such as façades and cars). The overall accuracy is increased by 7.01% on the Vienna dataset and 6.88% on the Vaihingen dataset. Moreover, most large, wrongly labeled regions are corrected by long-range interactions that are derived from the optimal graph, and misclassified regions that lack neighborhood communications in terms of correct labels are also corrected with the probabilistic label relaxation.〈/p〉〈/div〉 〈/div〉

Description: 〈p〉Publication date: January 2019〈/p〉 〈p〉〈b〉Source:〈/b〉 ISPRS Journal of Photogrammetry and Remote Sensing, Volume 147〈/p〉 〈p〉Author(s): Timo P. Pitkänen, Pasi Raumonen, Annika Kangas〈/p〉 〈div xml:lang="en"〉 〈h5〉Abstract〈/h5〉 〈div〉〈p〉Point clouds generated by terrestrial laser scanners (TLS) have enabled new ways to measure stem diameters. A common method for diameter calculation is to fit cylindrical or circular shapes into the TLS point cloud, which can be based either on a single scan or a co-registered combination of several scans. However, as various defects in the point cloud may affect the final diameter results, we propose an automatized processing chain which takes advantage of complementing steps. Processing consists of two fitting phases and an additional taper curve calculation to define the final diameter measurements. First, stems are detected from co-registered data of several scans using surface normals and cylinder fitting. This provides a robust framework for localizing the stems and estimating diameters at various heights. Then, guided by the cylinders and their indicative diameters, another fitting round is performed by cutting the stems into thin horizontal slices and reassessing their diameters by circular shape. For each slice, the quality of the cylinder-modelled diameter is evaluated first with co-registered data and if it is found to be deficient, potentially due to modelling defects or co-registration errors, diameter is detected through single scans. Finally, slice diameters are applied to construct a spline-based taper curve model for each tree, which is used to calculate the final stem dimensions. This methodology was tested in southern Finland using a set of 505 trees. At the breast height level (1.3 m), the results indicate 5.2 mm mean difference (3.2%), −0.4 mm bias (-0.3%) and 7.3 mm root mean squared error (4.4%) to reference measurements, and at the height of 6.0 m, respective values are 6.5 mm (3.6%), +1.6 mm (0.9%) and 8.4 mm (4.8%). These values are smaller compared to most of the corresponding contemporary studies, and outperform the initial cylinder models. This indicates that the applied processing chain is capable of producing relatively accurate diameter measurements, which can, at the cost of computational heaviness, remove various defects and improve the modelling results.〈/p〉〈/div〉 〈/div〉

Description: 〈p〉Publication date: January 2019〈/p〉 〈p〉〈b〉Source:〈/b〉 ISPRS Journal of Photogrammetry and Remote Sensing, Volume 147〈/p〉 〈p〉Author(s): Annalisa Appice, Donato Malerba〈/p〉 〈div xml:lang="en"〉 〈h5〉Abstract〈/h5〉 〈div〉〈p〉Classifying every pixel of a hyperspectral image with a certain land-cover type is the cornerstone of hyperspectral image analysis. In the present study a segmentation-aided methodology for the spectral-spatial classification of hyperspectral data is proposed. It considers the spatial dependence of the spectral bands, deals with the curse of dimensionality and handles the spectral variability. A local spatial regularization of spectral information is used, in order to derive an informative joint spectral-spatial representation of the data. A contiguity-based segmentation algorithm is formulated, in order to build the object-wise texture that can aid classifier learning. The hybrid use of the segmentation texture is evaluated in both pre-processing (i.e. selecting representative pixels to learn the classifier) and post-processing (i.e. refining predicted labels and removing possible outlier classifications). The experiments performed with the proposed methodology provide encouraging results, also compared to several recent state-of-the-art approaches.〈/p〉〈/div〉 〈/div〉

Description: 〈p〉Publication date: December 2019〈/p〉 〈p〉〈b〉Source:〈/b〉 ISPRS Journal of Photogrammetry and Remote Sensing, Volume 158〈/p〉 〈p〉Author(s): Dino Ienco, Roberto Interdonato, Raffaele Gaetano, Dinh Ho Tong Minh〈/p〉 〈div xml:lang="en"〉 〈h5〉Abstract〈/h5〉 〈div〉〈p〉The huge amount of data currently produced by modern Earth Observation (EO) missions has allowed for the design of advanced machine learning techniques able to support complex Land Use/Land Cover (LULC) mapping tasks. The Copernicus programme developed by the European Space Agency provides, with missions such as Sentinel-1 (S1) and Sentinel-2 (S2), radar and optical (multi-spectral) imagery, respectively, at 10 m spatial resolution with revisit time around 5 days. Such high temporal resolution allows to collect Satellite Image Time Series (SITS) that support a plethora of Earth surface monitoring tasks. How to effectively combine the complementary information provided by such sensors remains an open problem in the remote sensing field. In this work, we propose a deep learning architecture to combine information coming from S1 and S2 time series, namely TWINNS (TWIn Neural Networks for Sentinel data), able to discover spatial and temporal dependencies in both types of SITS. The proposed architecture is devised to boost the land cover classification task by leveraging two levels of complementarity, i.e., the interplay between radar and optical SITS as well as the synergy between spatial and temporal dependencies. Experiments carried out on two study sites characterized by different land cover characteristics (i.e., the 〈em〉Koumbia〈/em〉 site in Burkina Faso and 〈em〉Reunion Island〈/em〉, a overseas department of France in the Indian Ocean), demonstrate the significance of our proposal.〈/p〉〈/div〉 〈/div〉

Description: 〈p〉Publication date: December 2019〈/p〉 〈p〉〈b〉Source:〈/b〉 ISPRS Journal of Photogrammetry and Remote Sensing, Volume 158〈/p〉 〈p〉Author(s): Wojciech Gruszczyński, Edyta Puniach, Paweł Ćwiąkała, Wojciech Matwij〈/p〉 〈div xml:lang="en"〉 〈h5〉Abstract〈/h5〉 〈div〉〈p〉The main advantage of using unmanned aerial vehicles (UAVs) is the relatively low cost of collecting data, especially when using photogrammetry on images of relatively small areas. Additionally, they have high operational flexibility and the results have a high spatial and temporal resolution. To further facilitate the use of UAVs in photogrammetry, we developed an algorithm to filter out points that indicate areas covered in low vegetation (grass, crops) from the generated point cloud. This paper presents a three-layer filtering algorithm based on convolutional neural networks (CNNs) created for this specific purpose. The modular structure of the algorithm makes it easy to expand on and improve. The proposed solution allows errors in the height of digital elevation model (DEM) points caused by the influence of vegetation to be reduced by as much as 60–70% in relation to height errors from the raw data of high grass. At the same time, the solution presented here is practical for low grass because it does not weaken the model. The algorithm significantly reduces the errors in the DEM, as well as the products derived from the DEM.〈/p〉〈/div〉 〈/div〉

Description: 〈p〉Publication date: November 2019〈/p〉 〈p〉〈b〉Source:〈/b〉 ISPRS Journal of Photogrammetry and Remote Sensing, Volume 157〈/p〉 〈p〉Author(s): Yong Chen, Wei He, Naoto Yokoya, Ting-Zhu Huang〈/p〉 〈div xml:lang="en"〉 〈h5〉Abstract〈/h5〉 〈div〉〈p〉Cloud and cloud shadow (cloud/shadow) removal from multitemporal satellite images is a challenging task and has elicited much attention for subsequent information extraction. Regarding cloud/shadow areas as missing information, low-rank matrix/tensor completion based methods are popular to recover information undergoing cloud/shadow degradation. However, existing methods required to determine the cloud/shadow locations in advance and failed to completely use the latent information in cloud/shadow areas. In this study, we propose a blind cloud/shadow removal method for time-series remote sensing images by unifying cloud/shadow detection and removal together. First, we decompose the degraded image into low-rank clean image (surface-reflected) component and sparse (cloud/shadow) component, which can simultaneously and completely use the underlying characteristics of these two components. Meanwhile, the spatial-spectral total variation regularization is introduced to promote the spatial-spectral continuity of the cloud/shadow component. Second, the cloud/shadow locations are detected from the sparse component using a threshold method. Finally, we adopt the cloud/shadow detection results to guide the information compensation from the original observed images to better preserve the information in cloud/shadow-free locations. The problem of the proposed model is efficiently addressed using the alternating direction method of multipliers. Both simulated and real datasets are performed to demonstrate the effectiveness of our method for cloud/shadow detection and removal when compared with other state-of-the-art methods.〈/p〉〈/div〉 〈/div〉

Description: 〈p〉Publication date: November 2019〈/p〉 〈p〉〈b〉Source:〈/b〉 ISPRS Journal of Photogrammetry and Remote Sensing, Volume 157〈/p〉 〈p〉Author(s): Tawanda W. Gara, Roshanak Darvishzadeh, Andrew K. Skidmore, Tiejun Wang, Marco Heurich〈/p〉 〈div xml:lang="en"〉 〈h5〉Abstract〈/h5〉 〈div〉〈p〉Leaf traits at canopy level (〈em〉hereinafter〈/em〉 canopy traits) are conventionally expressed as a product of total canopy leaf area index (LAI) and leaf trait content based on samples collected from the exposed upper canopy. This traditional expression is centered on the theory that absorption of incident photosynthetically active radiation (PAR) follow a bell-shaped function skewed to the upper canopy. However, the validity of this theory has remained untested for a suite of canopy traits in a temperate forest ecosystem across multiple seasons using multispectral imagery. In this study, we examined the effect of canopy traits expression in modelling canopy traits using Sentinel-2 multispectral data across the growing season in Bavaria Forest National Park (BFNP), Germany. To achieve this, we measured leaf mass per area (LMA), chlorophyll (C〈sub〉ab〈/sub〉), nitrogen (N) and carbon content and LAI from the exposed upper and shaded lower canopy respectively over three seasons (spring, summer and autumn). Subsequently, we estimated canopy traits using two expressions, i.e. the traditional expression-based on the product of LAI and leaf traits content of samples collected from the sunlit upper canopy (〈em〉hereinafter〈/em〉 top-of-canopy expression) and the weighted expression - established on the proportion between the shaded lower and sunlit upper canopy LAI and their respective leaf traits content. Using a Random Forest machine-learning algorithm, we separately modelled canopy traits estimated from the two expressions using Sentinel-2 spectral bands and vegetation indices. Our results showed that dry matter related canopy traits (LMA, N and carbon) estimated based on the weighted canopy expression yield stronger correlations and higher prediction accuracy (NRMSE〈sub〉CV〈/sub〉 〈 0.19) compared to the top-of-canopy traits expression across all seasons. In contrast, canopy chlorophyll estimated from the top-of-canopy expression demonstrated strong fidelity with Sentinel-2 bands and vegetation indices (RMSE 〈 0.48 µg/cm〈sup〉2〈/sup〉) compared to weighted canopy chlorophyll (RMSE 〉 0.48 µg/cm〈sup〉2〈/sup〉) across all seasons. We also developed a generalized model that explained 52.57–67.82% variation in canopy traits across the three seasons. Using the most accurate Random Forest model for each season, we demonstrated the capability of Sentinel-2 data to map seasonal dynamics of canopy traits across the park. Results presented in this study revealed that canopy trait expression can have a profound effect on modelling the accuracy of canopy traits using satellite imagery throughout the growing seasons. These findings have implications on model accuracy when monitoring the dynamics of ecosystem functions, processes and services.〈/p〉〈/div〉 〈/div〉

Description: 〈p〉Publication date: December 2019〈/p〉 〈p〉〈b〉Source:〈/b〉 ISPRS Journal of Photogrammetry and Remote Sensing, Volume 158〈/p〉 〈p〉Author(s): Lin Du, Xiong You, Ke Li, Liqiu Meng, Gong Cheng, Liyang Xiong, Guangxia Wang〈/p〉 〈div xml:lang="en"〉 〈h5〉Abstract〈/h5〉 〈div〉〈p〉Automatic landform recognition is considered to be one of the most important tools for landform classification and deepening our understanding of terrain morphology. This paper presents a multi-modal geomorphological data fusion framework which uses deep learning-based methods to improve the performance of landform recognition. It leverages a multi-channel geomorphological feature extraction network to generate different characteristics from multi-modal geomorphological data, such as shaded relief, DEM, and slope and then it harvests joint features via a multi-modal geomorphological feature fusion network in order to effectively represent landforms. A residual learning unit is used to mine deep correlations from visual and physical modality features to achieve the final landform representations. Finally, it employs three fully-connected layers and a softmax classifier to generate labels for each sample data. Experimental results indicate that this multi-modal data fusion-based algorithm obtains much better performance than conventional algorithms. The highest recognition rate was 90.28%, showing a great potential for landform recognition.〈/p〉〈/div〉 〈/div〉

Description: 〈p〉Publication date: December 2019〈/p〉 〈p〉〈b〉Source:〈/b〉 ISPRS Journal of Photogrammetry and Remote Sensing, Volume 158〈/p〉 〈p〉Author(s): Hamid Hamraz, Nathan B. Jacobs, Marco A. Contreras, Chase H. Clark〈/p〉 〈div xml:lang="en"〉 〈h5〉Abstract〈/h5〉 〈div〉〈p〉The purpose of this study was to investigate the use of deep learning for coniferous/deciduous classification of individual trees segmented from airborne LiDAR data. To enable processing by a deep convolutional neural network (CNN), we designed two discrete representations using leaf-off and leaf-on LiDAR data: a digital surface model with four channels (DSM × 4) and a set of four 2D views (4 × 2D). A training dataset of tree crowns was generated via segmentation of tree crowns, followed by co-registration with field data. Potential mislabels due to GPS error or tree leaning were corrected using a statistical ensemble filtering procedure. Because the training data was heavily unbalanced (~8% conifers), we trained an ensemble of CNNs on random balanced sub-samples. Benchmarked against multiple traditional shallow learning methods using manually designed features, the CNNs improved accuracies up to 14%. The 4 × 2D representation yielded similar classification accuracies to the DSM × 4 representation (~82% coniferous and ~90% deciduous) while converging faster. Further experimentation showed that early/late fusion of the channels in the representations did not affect the accuracies in a significant way. The data augmentation that was used for the CNN training improved the classification accuracies, but more real training instances (especially coniferous) likely results in much stronger improvements. Leaf-off LiDAR data were the primary source of useful information, which is likely due to the perennial nature of coniferous foliage. LiDAR intensity values also proved to be useful, but normalization yielded no significant improvement. As we observed, large training data may compensate for the lack of a subset of important domain data. Lastly, the classification accuracies of overstory trees (~90%) were more balanced than those of understory trees (~90% deciduous and ~65% coniferous), which is likely due to the incomplete capture of understory tree crowns via airborne LiDAR. In domains like remote sensing and biomedical imaging, where the data contain a large amount of information and are not friendly to human visual system, human-designed features may become suboptimal. As exemplified by this study, automatic, objective derivation of optimal features via deep learning can improve prediction tasks in such domains.〈/p〉〈/div〉 〈/div〉

Description: 〈p〉Publication date: May 2019〈/p〉 〈p〉〈b〉Source:〈/b〉 ISPRS Journal of Photogrammetry and Remote Sensing, Volume 151〈/p〉 〈p〉Author(s): Christian Geiß, Patrick Aravena Pelizari, Lukas Blickensdörfer, Hannes Taubenböck〈/p〉 〈div xml:lang="en"〉 〈h5〉Abstract〈/h5〉 〈div〉〈p〉We follow the idea of learning invariant decision functions for remote sensing image classification with Support Vector Machines (SVM). To do so, we generate artificially transformed samples (i.e., virtual samples) from available prior knowledge. Labeled samples closest to the separating hyperplane with maximum margin (i.e., the Support Vectors) are identified by learning an initial SVM model. The Support Vectors are used for generating virtual samples by perturbing the features to which the model should be invariant. Subsequently, the model is relearned using the Support Vectors and the virtual samples to eventually alter the hyperplane with maximum margin and enhance generalization capabilities of decision functions. In contrast to existing approaches, we establish a self-learning procedure to ultimately prune non-informative virtual samples from a possibly arbitrary invariance generation process to allow for robust and sparse model solutions. The self-learning strategy jointly considers a similarity and margin sampling constraint. In addition, we innovatively explore the invariance generation process in the context of an object-based image analysis framework. Image elements (i.e., pixels) are aggregated to image objects (as represented by segments/superpixels) with a segmentation algorithm. From an initial singular segmentation level, invariances are encoded by varying hyperparameters of the segmentation algorithm in terms of scale and shape. Experimental results are obtained from two very high spatial resolution multispectral data sets acquired over the city of Cologne, Germany, and the Hagadera Refugee Camp, Kenya. Comparative model accuracy evaluations underline the favorable performance properties of the proposed methods especially in settings with very few labeled samples.〈/p〉〈/div〉 〈/div〉

Description: 〈p〉Publication date: May 2019〈/p〉 〈p〉〈b〉Source:〈/b〉 ISPRS Journal of Photogrammetry and Remote Sensing, Volume 151〈/p〉 〈p〉Author(s): Ruitao Feng, Qingyun Du, Xinghua Li, Huanfeng Shen〈/p〉 〈div xml:lang="en"〉 〈h5〉Abstract〈/h5〉 〈div〉〈p〉Highly accurate registration is one of the essential requirements for numerous applications of remote sensing images. Toward this end, we have developed a robust algorithm by combining and localizing feature- and area-based methods. A block-weighted projective (BWP) transformation model is first employed to map the local geometric relationship with weighted feature points in the feature-based stage, for which the weight is determined by an inverse distance weighted (IDW) function. Subsequently, the outlier-insensitive (OIS) model aims to further optimize the registration in the area-based stage. Considering the inevitable outliers (e.g., cloud, noise, land-cover change), OIS integrates Huber estimation with the structure tensor (ST), which is an approach that is robust to residual errors and outliers while preserving edges. Four pairs of remote sensing images with varied terrain features were tested in the experiments. Compared with the-state-of-art algorithms, the proposed algorithm is more effective, in terms of both visual quality and quantitative evaluation.〈/p〉〈/div〉 〈/div〉

Description: 〈p〉Publication date: May 2019〈/p〉 〈p〉〈b〉Source:〈/b〉 ISPRS Journal of Photogrammetry and Remote Sensing, Volume 151〈/p〉 〈p〉Author(s): Maitiniyazi Maimaitijiang, Vasit Sagan, Paheding Sidike, Matthew Maimaitiyiming, Sean Hartling, Kyle T. Peterson, Michael J.W. Maw, Nadia Shakoor, Todd Mockler, Felix B. Fritschi〈/p〉 〈div xml:lang="en"〉 〈h5〉Abstract〈/h5〉 〈div〉〈p〉Crop biomass estimation with high accuracy at low-cost is valuable for precision agriculture and high-throughput phenotyping. Recent technological advances in Unmanned Aerial Systems (UAS) significantly facilitate data acquisition at low-cost along with high spatial, spectral, and temporal resolution. The objective of this study was to explore the potential of UAS RGB imagery-derived spectral, structural, and volumetric information, as well as a proposed vegetation index weighted canopy volume model (CVM〈sub〉VI〈/sub〉) for soybean [〈em〉Glycine〈/em〉 max (L.) Merr.] aboveground biomass (AGB) estimation. RGB images were collected from low-cost UAS throughout the growing season at a field site near Columbia, Missouri, USA. High-density point clouds were produced using the structure from motion (SfM) technique through a photogrammetric workflow based on UAS stereo images. Two-dimensional (2D) canopy structure metrics such as canopy height (CH) and canopy projected basal area (BA), as well as three-dimensional (3D) volumetric metrics such as canopy volume model (CVM) were derived from photogrammetric point clouds. A variety of vegetation indices (VIs) were also extracted from RGB orthomosaics. Then, CVM〈sub〉VI〈/sub〉, which combines canopy spectral and volumetric information, was proposed. Commonly used regression models were established based on the UAS-derived information and field- measured AGB with a leave-one-out cross-validation. The results show that: (1) In general, canopy 2D structural metrics CH and BA yielded higher correlation with AGB than VIs. (2) Three-dimensional metrics, such as CVM, that encompass both horizontal and vertical properties of canopy provided better estimates for AGB compared to 2D structural metrics (R〈sup〉2〈/sup〉 = 0.849; RRMSE = 18.7%; MPSE = 20.8%). (3) Optimized CVM〈sub〉VI〈/sub〉, which incorporates both canopy spectral and 3D volumetric information outperformed the other indices and metrics, and was a better predictor for AGB estimation (R〈sup〉2〈/sup〉 = 0.893; RRMSE = 16.3%; MPSE = 19.5%). In addition, CVM〈sub〉VI〈/sub〉 showed equal prediction power for different genotypes, which indicates its potential for high-throughput soybean biomass estimation. Moreover, a CVM〈sub〉VI〈/sub〉 based univariate regression model yielded AGB predicting capability comparable to multivariate complex regression models such as stepwise multilinear regression (SMR) and partial least squares regression (PLSR) that incorporate multiple canopy spectral indices and structural metrics. Overall, this study reveals the potential of canopy spectral, structural and volumetric information, and their combination (i.e., CVM〈sub〉VI〈/sub〉) for estimations of soybean AGB. CVM〈sub〉VI〈/sub〉 was shown to be simple but effective in estimating AGB, and could be applied for high-throughput phenotyping and precision agro-ecological applications and management.〈/p〉〈/div〉 〈/div〉

Description: 〈p〉Publication date: May 2019〈/p〉 〈p〉〈b〉Source:〈/b〉 ISPRS Journal of Photogrammetry and Remote Sensing, Volume 151〈/p〉 〈p〉Author(s): Bisheng Yang, Zhen Dong, Yuan Liu, Fuxun Liang, Yongjun Wang〈/p〉

Description: 〈p〉Publication date: May 2019〈/p〉 〈p〉〈b〉Source:〈/b〉 ISPRS Journal of Photogrammetry and Remote Sensing, Volume 151〈/p〉 〈p〉Author(s): Benqing Chen, Yanming Yang, Dewei Xu, Erhui Huang〈/p〉 〈div xml:lang="en"〉 〈h5〉Abstract〈/h5〉 〈div〉〈p〉For shallow water depth retrieval from high spatial resolution satellite images, although numerous empirical models have been developed, it remains impossible to estimate shallow water depths without collection of required ground truth depth. To address this limitation, a new physically based dual band algorithm is developed to estimate shallow water depths using blue and green bands from high spatial resolution multispectral image with no ground truth. The dual band log-linear model is first analytically formulated, which then is used for shallow water depths retrieval by solving all unknown model parameters based on different types of sampling pixels directly extracted from the multispectral image. The adjacent pixel pairs from the intersecting edges of different bottom types across various depths over shallow water area, are employed to calculate the optimal band rotation coefficient unit vector by minimization method. On the basis, the bottom parameter is estimated through the pixels from the coastline. Additionally, the pixels from various depths of same bottom type are also employed to achieve the blue to green band ratio of diffused attenuation coefficient. The sum of the diffuse attenuation coefficients of green band for upwelling and downwelling light is estimated by QAA and Kd algorithms. To evaluate the performance of the proposed algorithm, the GeoEye-1 image covered Jinqing Island and the Chinese Gaofen-2 image across Kaneohe Bay are chosen to achieve shallow water depth by using the proposed algorithm after geo-rectification and atmospheric correction. The validations using the actual water depths show the overall root mean square errors (RMSEs) for the derived water depths are 1.18 m for Jinqing Island and 1.34 m for Kaneohe Bay respectively. Compared to the Lyzenga empirical model, the developed approach can generally achieve slightly better results for shallow water depths with no ground truth data. Finally, the effects of the variation in the model parameters to water depth retrieval are discussed and analyzed.〈/p〉〈/div〉 〈/div〉

Description: 〈p〉Publication date: May 2019〈/p〉 〈p〉〈b〉Source:〈/b〉 ISPRS Journal of Photogrammetry and Remote Sensing, Volume 151〈/p〉 〈p〉Author(s): Fatma Haouas, Basel Solaiman, Zouhour Ben Dhiaf, Atef Hamouda, Khaled Bsaies〈/p〉 〈div xml:lang="en"〉 〈h5〉Abstract〈/h5〉 〈div〉〈p〉Change detection monitoring on multi-temporal remote sensed images is a persistent methodological challenge where the Dempster-Shafer, or evidence, Theory (DST) has been often applied. This paper presents a new method based on the use of DST for mining bi-temporal remotely sensed images change. The main idea is based on the investigation, analysis and interpretation of different types of conflict between two bi-temporal mass distributions. The reasoning process is focused on the conflict significance and its “partial” causes. In fact, the global conflict that occurs during the joint exploitation of multi-temporal images gives general and non-sufficiently concise information. However, the partial conflict provides rich and important information with regards to the disagreement between knowledge sources. For computing the partial conflict between focal elements, the geometric representation of mass distributions is exploited. The obtained conflict measures, caused by change, are analyzed latter by a new algorithm for drifting binary change map and identifying change directions. The effectiveness and reliability of the proposed approach are shown through experimentations on simulated changed images as well as using multi-temporal Landsat satellite images where qualitative criteria as well as quantitative measures are applied. The performances of the proposed approach, in terms of changed area recognition, are compared to three different and widely used conflict measures: the Empty-set mass, the Jousselme’s distance and the Cosine measure. It is shown that the developed change detection approach outperforms these conflict measures.〈/p〉〈/div〉 〈/div〉

Description: 〈p〉Publication date: April 2019〈/p〉 〈p〉〈b〉Source:〈/b〉 ISPRS Journal of Photogrammetry and Remote Sensing, Volume 150〈/p〉 〈p〉Author(s): 〈/p〉

Description: 〈p〉Publication date: May 2019〈/p〉 〈p〉〈b〉Source:〈/b〉 ISPRS Journal of Photogrammetry and Remote Sensing, Volume 151〈/p〉 〈p〉Author(s): 〈/p〉

Description: 〈p〉Publication date: June 2019〈/p〉 〈p〉〈b〉Source:〈/b〉 ISPRS Journal of Photogrammetry and Remote Sensing, Volume 152〈/p〉 〈p〉Author(s): Dalin Jiang, Bunkei Matsushita, Fajar Setiawan, Augusto Vundo〈/p〉 〈div xml:lang="en"〉 〈h5〉Abstract〈/h5〉 〈div〉〈p〉The Secchi disk depth (Z〈sub〉SD〈/sub〉) is a widely used parameter for evaluating water clarity. Here we propose an improved algorithm, which is based on a new underwater visibility theory, for retrieving more accurate Z〈sub〉SD〈/sub〉 from remote sensing reflectance (〈em〉R〈sub〉rs〈/sub〉〈/em〉) in various waters. Two improvements were carried out in the new algorithm. First, we used a hybrid quasi-analytical algorithm (QAA_hybrid) instead of the sixth version of QAA (QAA_v6) for retrieving more accurate total absorption coefficient (〈math xmlns:mml="http://www.w3.org/1998/Math/MathML" altimg="si2.gif" overflow="scroll"〉〈mrow〉〈mi〉a〈/mi〉〈mrow〉〈mfenced open="(" close=")"〉〈mrow〉〈mi〉λ〈/mi〉〈/mrow〉〈/mfenced〉〈/mrow〉〈/mrow〉〈/math〉) and total backscattering coefficient (〈math xmlns:mml="http://www.w3.org/1998/Math/MathML" altimg="si3.gif" overflow="scroll"〉〈mrow〉〈msub〉〈mi〉b〈/mi〉〈mi〉b〈/mi〉〈/msub〉〈mrow〉〈mfenced open="(" close=")"〉〈mrow〉〈mi〉λ〈/mi〉〈/mrow〉〈/mfenced〉〈/mrow〉〈/mrow〉〈/math〉) even in turbid inland waters. Second, we used a dynamic 〈math xmlns:mml="http://www.w3.org/1998/Math/MathML" altimg="si4.gif" overflow="scroll"〉〈mrow〉〈msub〉〈mi〉K〈/mi〉〈mi〉T〈/mi〉〈/msub〉〈mo stretchy="false"〉/〈/mo〉〈msub〉〈mi〉K〈/mi〉〈mi〉d〈/mi〉〈/msub〉〈/mrow〉〈/math〉 ratio (i.e., ratio of diffuse attenuation coefficient of upwelling radiance and diffuse attenuation coefficient of downwelling irradiance) instead of using the fixed ratio (i.e., 1.5). The results obtained from in situ 〈em〉R〈sub〉rs〈/sub〉〈/em〉 show that the improved Z〈sub〉SD〈/sub〉 estimation algorithm gave more accurate Z〈sub〉SD〈/sub〉 estimations, with the root mean square error (RMSE) reduced from 0.2 to 0.1 in log10 unit, mean absolute percentage error (MAPE) reduced from 39% to 20% (N = 178 with in situ Z〈sub〉SD〈/sub〉 values between 0.3 and 20.8 m). We then applied the improved Z〈sub〉SD〈/sub〉 estimation algorithm to the 2003–2012 MERIS images for Lake Kasumigaura to further confirm the performance of the improved Z〈sub〉SD〈/sub〉 estimation algorithm. The results obtained from 19 matchups demonstrate that the estimated Z〈sub〉SD〈/sub〉 matched well with the in situ Z〈sub〉SD〈/sub〉, with the RMSE of 0.11 m and the MAPE of 15%. The improved Z〈sub〉SD〈/sub〉 estimation algorithm shows a potential to estimate more accurate Z〈sub〉SD〈/sub〉 values from remote sensing data in various waters.〈/p〉〈/div〉 〈/div〉

Description: 〈p〉Publication date: June 2019〈/p〉 〈p〉〈b〉Source:〈/b〉 ISPRS Journal of Photogrammetry and Remote Sensing, Volume 152〈/p〉 〈p〉Author(s): Sebastián Ortega, Agustín Trujillo, José Miguel Santana, José Pablo Suárez, Jaisiel Santana〈/p〉 〈div xml:lang="en"〉 〈h5〉Abstract〈/h5〉 〈div〉〈p〉As the electric companies need to assure the reliability of their services, power line management gains importance during the last years. Many of them rely on LiDAR scanning of their assets to obtain the status of their power line corridors and determine possible risks. In this paper, a novel sevenfold staged pipeline is introduced to classify pylon and wire points and model the conductors. Wire points are subdivided into three categories: shield, common conductor and chain. Pylons of two different types are taken into account: suspension and anchor. For the first case, insulator strains are also identified and separated. Wire points are segmented as individual conductors and a 3D-wise model based on the catenary equation is generated for each conductor using particle swarm optimization. Tests have been conducted on a set with 25 point cloud files to assess the accuracy and correctness of the results given by the proposed pipeline.〈/p〉〈/div〉 〈/div〉 〈div xml:lang="en"〉 〈h5〉Graphical abstract〈/h5〉 〈div〉〈p〉〈figure〉〈img src="https://ars.els-cdn.com/content/image/1-s2.0-S0924271619300905-ga1.jpg" width="444" alt="Graphical abstract for this article" title=""〉〈/figure〉〈/p〉〈/div〉 〈/div〉

Description: 〈p〉Publication date: August 2019〈/p〉 〈p〉〈b〉Source:〈/b〉 ISPRS Journal of Photogrammetry and Remote Sensing, Volume 154〈/p〉 〈p〉Author(s): Ali Mokhtari, Hamideh Noory, Farrokh Pourshakouri, Parisa Haghighatmehr, Yasamin Afrasiabian, Maryam Razavi, Fatemeh Fereydooni, Ali Sadeghi Naeni〈/p〉 〈div xml:lang="en"〉 〈h5〉Abstract〈/h5〉 〈div〉〈p〉Evapotranspiration is considered to be an important component of allocating water to agricultural sector; therefore, the more accurate this parameter is, the more optimized the water use can be. This study was conducted in order to evaluate the Landsat 8 and Sentinel-2 data (A and B), both separately and combined, in potential evapotranspiration (ET〈sub〉p〈/sub〉) and single crop coefficient (K〈sub〉c〈/sub〉) estimations. Field measurements such as crop height, leaf area index (LAI), land surface temperature (LST), air temperature above canopy (AT), and spectral data were exploited in the evaluating process throughout the entirety of 2017–18 growing season under winter wheat and barley cultivations in the Agricultural Research Farms of the University of Tehran. The novel method of Multi-Sensor Data Fusion using the Priestly-Taylor equation was taken into practice for satellite-based ET〈sub〉p〈/sub〉 (MSDF-ET) calculation from the combination of MODIS thermal and Landsat 8 and Sentinel-2 multispectral data. Thermal images were downscaled by the means of the TsHARP algorithm. Thus, prior to ET〈sub〉p〈/sub〉 calculation, the thermal sharpening algorithm calculated using different spectral indices (SI) was assessed. The SI included NDVI, SAVI, SR, NDWI, NDWI〈sub〉g〈/sub〉, and LSWI. The subsequent results were representative of the LSWI qualification under both Landsat 8 and Sentinel-2 conditions against thermal and spectral measurements. Also the satellite-based ET〈sub〉p〈/sub〉 strongly correlated with the ET〈sub〉p〈/sub〉 derived from the field data illuminating the promising accuracy of the MSDF-ET method in both Landsat 8 and Sentinel-2 data. In the end, the time series of K〈sub〉c〈/sub〉 obtained from the combination of satellites were fairly indicative of the real-world variations under different vegetation cover and crop growth stages. Overall, using Landsat 8 and Sentinel-2 products in integration with each other could significantly result in more reliable decisions in agricultural water resources management.〈/p〉〈/div〉 〈/div〉

Description: 〈p〉Publication date: June 2019〈/p〉 〈p〉〈b〉Source:〈/b〉 ISPRS Journal of Photogrammetry and Remote Sensing, Volume 152〈/p〉 〈p〉Author(s): Lei Zheng, Guosong Zhao, Jinwei Dong, Quansheng Ge, Jian Tao, Xuezhen Zhang, Youcun Qi, Russell B. Doughty, Xiangming Xiao〈/p〉 〈div xml:lang="en"〉 〈h5〉Abstract〈/h5〉 〈div〉〈p〉Changes in Earth’s albedo due to vegetation dynamics, snow cover, and land cover change have attracted much attention. However, the effects of vegetation dynamics on albedo have not been comprehensively documented according to its spatial (regional), temporal (within growing season), and spectral (visible, near-infrared, and shortwave) characteristics. This study examined the effects of vegetation greenness on albedo from 2000 to 2014 in China’s grasslands, which have considerable intra- and inter-annual variations, using remote sensing-based albedo and two-band Enhanced Vegetation Index (EVI2) data. Generally, we found an insignificant negative correlation between the shortwave (SW) albedo and EVI2 for grasslands in China. However, the visible (VIS) albedo was more sensitive to changes in vegetation greenness than near-infrared (NIR) albedo in China’s grasslands. The relationship between the NIR albedo and EVI2 was more complicated, especially in the Tibetan Plateau (TP), where the correlation was negative in the early growing season and positive in the late growing season; while the correlation between the NIR albedo and EVI2 was always negative in main part of Inner Mongolia (IM). The different albedo-EVI2 relationships in IM and TP may be related to differences in soil albedos. The higher sensitivity of the SW albedo to vegetation greenness change in IM, the stronger effect on land surface radiation budget. Our finding about vegetation-induced changes in albedo differ in space, time and spectral bands is expected to contribute to the improvement of land surface models.〈/p〉〈/div〉 〈/div〉

Description: 〈p〉Publication date: August 2019〈/p〉 〈p〉〈b〉Source:〈/b〉 ISPRS Journal of Photogrammetry and Remote Sensing, Volume 154〈/p〉 〈p〉Author(s): Jie Wang, Xiangming Xiao, Rajen Bajgain, Patrick Starks, Jean Steiner, Russell B. Doughty, Qing Chang〈/p〉 〈div xml:lang="en"〉 〈h5〉Abstract〈/h5〉 〈div〉〈p〉Grassland degradation has accelerated in recent decades in response to increased climate variability and human activity. Rangeland and grassland conditions directly affect forage quality, livestock production, and regional grassland resources. In this study, we examined the potential of integrating synthetic aperture radar (SAR, Sentinel-1) and optical remote sensing (Landsat-8 and Sentinel-2) data to monitor the conditions of a native pasture and an introduced pasture in Oklahoma, USA. Leaf area index (LAI) and aboveground biomass (AGB) were used as indicators of pasture conditions under varying climate and human activities. We estimated the seasonal dynamics of LAI and AGB using Sentinel-1 (S1), Landsat-8 (LC8), and Sentinel-2 (S2) data, both individually and integrally, applying three widely used algorithms: Multiple Linear Regression (MLR), Support Vector Machine (SVM), and Random Forest (RF). Results indicated that integration of LC8 and S2 data provided sufficient data to capture the seasonal dynamics of grasslands at a 10–30-m spatial resolution and improved assessments of critical phenology stages in both pluvial and dry years. The satellite-based LAI and AGB models developed from ground measurements in 2015 reasonably predicted the seasonal dynamics and spatial heterogeneity of LAI and AGB in 2016. By comparison, the integration of S1, LC8, and S2 has the potential to improve the estimation of LAI and AGB more than 30% relative to the performance of S1 at low vegetation cover (LAI 〈 2 m〈sup〉2〈/sup〉/m〈sup〉2〈/sup〉, AGB 〈 500 g/m〈sup〉2〈/sup〉) and optical data of LC8 and S2 at high vegetation cover (LAI 〉 2 m〈sup〉2〈/sup〉/m〈sup〉2〈/sup〉, AGB 〉 500 g/m〈sup〉2〈/sup〉). These results demonstrate the potential of combining S1, LC8, and S2 monitoring grazing tallgrass prairie to provide timely and accurate data for grassland management.〈/p〉〈/div〉 〈/div〉

Description: 〈p〉Publication date: April 2019〈/p〉 〈p〉〈b〉Source:〈/b〉 ISPRS Journal of Photogrammetry and Remote Sensing, Volume 150〈/p〉 〈p〉Author(s): Yanbiao Sun, Stuart Robson, Daniel Scott, Jan Boehm, Qiang Wang〈/p〉 〈div xml:lang="en"〉 〈h5〉Abstract〈/h5〉 〈div〉 〈p〉To improve the accuracy of sensor orientation using calibrated aerial images, this paper proposes an automatic sensor orientation method utilizing horizontal and vertical constraints on human-engineered structures, addressing the limitations faced with sub-optimal number of Ground Control Points (GCPs) within a scene. Related state-of-the-art methods rely on structured building edges, and necessitate manual identification of end points. Our method makes use of line-segments but eliminates the need for these matched end points, thus eliminating the need for inefficient manual intervention.〈/p〉 〈p〉To achieve this, a 3D line in object space is represented by the intersection of two planes going through two camera centers. The normal vector of each plane can be written as a function of a pair of azimuth and elevations angles. The normal vector of the 3D line can be expressed by the cross product of these two plane’s normal vectors. Then, we create observation functions of horizontal and vertical line constraints based on the zero-vector cross-product and the dot-product of the normal vector of the 3D lines. The observation functions of the horizontal and vertical lines are then introduced into a hybrid Bundle Adjustment (BA) method as constraints, including observed image points as well as observed line segment projections. Finally, to assess the feasibility and effectiveness of the proposed method, simulated and real data are tested. The results demonstrate that, in cases with only 3 GCPs, the accuracy of the proposed method utilizing line features extracted automatically, is increased by 50%, compared to a BA using only point constraints.〈/p〉 〈/div〉 〈/div〉

Description: 〈p〉Publication date: April 2019〈/p〉 〈p〉〈b〉Source:〈/b〉 ISPRS Journal of Photogrammetry and Remote Sensing, Volume 150〈/p〉 〈p〉Author(s): X.Q. Xu, J.S. Lu, N. Zhang, T.C. Yang, J.Y. He, X. Yao, T. Cheng, Y. Zhu, W.X. Cao, Y.C. Tian〈/p〉 〈div xml:lang="en"〉 〈h5〉Abstract〈/h5〉 〈div〉〈p〉The radiative transfer model (RTM) simulates forward spectral reflectance of vegetation and is used to estimate physical parameters using backwards inversion. However, differentiation of spectral reflectances may be hampered due to model parameter combinations, and the cost function within RTM that calculates statistical distance may lead to inconsistent inversions. Bayesian network (BN) is a probabilistic model that is used to solve problems of model ambiguity and incompleteness. Here, we constructed a model to estimate rice growth parameters using data collected by an unmanned aerial vehicle (UAV). We collected rice canopy spectral information using a MiniMCA-6 multispectral camera fitted to an UAV that was used to determine BN structure using parameters derived from the PROSAIL model. We calculated conditional probability distributions of different observed combinations of rice canopy chlorophyll content (CCC) and leaf area index (LAI) and a look up table of maximum conditional probabilities of rice growth parameters based on BN was developed. Results indicated that most accurate inversions of LAI and CCC as BN nodes were achieved at reflectances of 720 nm, 800 nm under the red normalized difference vegetation index and at reflectances of 550, 720, 800 nm under the modified simple ratio index, respectively. Compared with the cost function inversion method, the BN method mitigated the ill-posed problem of inversion and obtained higher inversion accuracy with model test R〈sup〉2〈/sup〉, RRMSE, and RE values of 0.81, 0.31, and 0.38, and 0.83, 0.36, and 0.43 for LAI and CCC, respectively. We conclude that application of the BN method to the inversion process of crop RTM could improve inversion accuracy of estimation of crop parameters.〈/p〉〈/div〉 〈/div〉

Description: 〈p〉Publication date: June 2019〈/p〉 〈p〉〈b〉Source:〈/b〉 ISPRS Journal of Photogrammetry and Remote Sensing, Volume 152〈/p〉 〈p〉Author(s): Yu Li, Sandro Martinis, Marc Wieland〈/p〉 〈div xml:lang="en"〉 〈h5〉Abstract〈/h5〉 〈div〉〈p〉Synthetic Aperture Radar (SAR) remote sensing has been widely used for flood mapping and monitoring. Nevertheless, flood detection in urban areas still proves to be particularly challenging by using SAR. In this paper, we assess the roles of SAR intensity and interferometric coherence in urban flood detection using multi-temporal TerraSAR-X data. We further introduce an active self-learning convolution neural network (A-SL CNN) framework to alleviate the effect of a limited annotated training dataset. The proposed framework selects informative unlabeled samples based on a temporal-ensembling CNN model. These samples are subsequently pseudo-labeled by a multi-scale spatial filter. Consistency regularization is introduced to penalize incorrect labels caused by pseudo-labeling. We show results for a case study that is centered on flooded areas in Houston, USA, during hurricane Harvey in August 2017. Our experiments show that multi-temporal intensity (pre- and co-event) plays the most important role in urban flood detection. Adding multi-temporal coherence can increase the reliability of the inundation map considerably. Meanwhile, encouraging results are achieved by the proposed A-SL CNN framework: the 〈em〉к〈/em〉 statistic is improved from 0.614 to 0.686 in comparison to its supervised counterpart.〈/p〉〈/div〉 〈/div〉

Description: 〈p〉Publication date: June 2019〈/p〉 〈p〉〈b〉Source:〈/b〉 ISPRS Journal of Photogrammetry and Remote Sensing, Volume 152〈/p〉 〈p〉Author(s): Bijan Seyednasrollah, Thomas Milliman, Andrew D. Richardson〈/p〉 〈div xml:lang="en"〉 〈h5〉Abstract〈/h5〉 〈div〉〈p〉Digital repeat photography and near-surface remote sensing have been used by environmental scientists to study environmental change for nearly a decade. However, a user-friendly, reliable, and robust platform to extract color-based statistics and time series from a large stack of images is still lacking. Here, we present an interactive open-source toolkit, called 〈em〉xROI〈/em〉, that facilitates the process of time series extraction and improves the quality of the final data. 〈em〉xROI〈/em〉 provides a responsive environment for scientists to interactively (a) delineate regions of interest (ROI), (b) handle field of view (FOV) shifts, and (c) extract and export time series data characterizing color-based metrics. The software gives user the opportunity to adjust mask files or draw new masks, every time an FOV shift occurs. Utilizing xROI can significantly facilitate data extraction from digital repeat photography and enhance the accuracy and continuity of extracted data.〈/p〉〈/div〉 〈/div〉

Description: 〈p〉Publication date: June 2019〈/p〉 〈p〉〈b〉Source:〈/b〉 ISPRS Journal of Photogrammetry and Remote Sensing, Volume 152〈/p〉 〈p〉Author(s): Wei Zhao, Hua Wu, Gaofei Yin, Si-Bo Duan〈/p〉 〈div xml:lang="en"〉 〈h5〉Abstract〈/h5〉 〈div〉〈p〉Information about land surface temperature (LST) acquired from remote sensing satellite observations is very important to monitor surface energy and water exchange processes at the land-atmosphere interface. However, the wide-view of the popularly used polar-orbiting satellites (Terra and Aqua) face the challenge from the temporal effect on their LST products induced by the big temporal differences along the scan line. To generate a time-consistent LST product, a practical normalization method is proposed in this study based on random forest regression for LST observations from the Moderate Resolution Imaging Spectroradiometer (MODIS) sensor onboard Terra satellite. A linking model is constructed to express LST as a function of various surface variables including vegetation indices, leaf area index, surface albedo, water index, solar radiation factor, and surface elevation. Under the assumption that the temporal effect is induced primarily by the differences in incident solar radiation, the temporal effect normalization is conducted by deriving a temporally consistent solar radiation factor which is used to drive the linking model and obtain the normalized LST. The proposed method is applied to the central Iberian Peninsula on the day of year 170 and 181, 2015. Results show that the areas with positive complement in solar radiation factor generally exhibit a positive increase in LST. An obvious improvement can be observed in the spatial pattern of the normalized LST data with the disappearance of the temperature boundary due to the big difference of satellite observation time. The Meteosat Second Generation (MSG) LST data which have the observations at the same local solar time is used for quantitative validation. The evaluation shows that the normalized LST data is more coincident with the MSG LST data than the original MODIS LST data, with significant improvements in the root mean squared deviation and bias with the MSG LST data (1.23 K and 1.66 K, respectively). Unlike previous normalization methods, the proposed method is conducted based on only satellite observations without other ancillary data. Therefore, the method demonstrates good potential for normalizing the temporal effect of the wide-view polar-orbiting satellite observations.〈/p〉〈/div〉 〈/div〉

Description: 〈p〉Publication date: June 2019〈/p〉 〈p〉〈b〉Source:〈/b〉 ISPRS Journal of Photogrammetry and Remote Sensing, Volume 152〈/p〉 〈p〉Author(s): Przemyslaw Polewski, Wei Yao〈/p〉 〈div xml:lang="en"〉 〈h5〉Abstract〈/h5〉 〈div〉〈p〉In this work, we investigate the coregistration of multimodal data, such as photogrammetric/LiDAR point clouds, digital surface models, orthoimages, or 3D CAD city models, using corresponding line segments. The lines are analytically derived as intersections of adjacent planar surfaces, which can be determined more robustly and are deemed more accurate compared to single point based features. We propose a two-stage approach, which first focuses on finding optimal line correspondences between the datasets using a scale-invariant graph matching method, and then utilizes the found matching as a basis for calculating the optimal coregistration transform. By decoupling the correspondence search from the transform calculation, our approach can use more line pairs for determining the optimal transform than would be practicable with a combined, sampling-style approach. As opposed to competing methods, our transform computation is based on explicitly minimizing the average L1 distance on the matched line set. The assumed model accounts for an isotropic scaling factor, three translations and three rotation angles. We conducted experiments on two publicly available ISPRS datasets: Vaihingen and Dortmund, and compared the performance of several variations of our approach with three competing methods. The results indicate that the L1 methods decreased the median matched line distance by up to one third in case of pre-aligned Z axes. Moreover, when coregistering two photogrammetric datasets acquired from distinct viewing perspectives, our method was able to triple the number of matched lines (under a strict proximity-based criterion) compared to its competitor. Our results show that it is worthwhile to base the transform calculation on significantly more line pairs than is customary for sample consensus-based approaches. Our established validation dataset for line-based coregistration has been published and made available online (〈a href="http://dx.doi.org/10.17632/dmp7tkn8kc.2" target="_blank"〉https://doi.org/10.17632/dmp7tkn8kc.2〈/a〉).〈/p〉〈/div〉 〈/div〉

Description: 〈p〉Publication date: April 2019〈/p〉 〈p〉〈b〉Source:〈/b〉 ISPRS Journal of Photogrammetry and Remote Sensing, Volume 150〈/p〉 〈p〉Author(s): Debaditya Acharya, Milad Ramezani, Kourosh Khoshelham, Stephan Winter〈/p〉 〈div xml:lang="en"〉 〈h5〉Abstract〈/h5〉 〈div〉〈p〉This article presents an accurate and robust visual indoor localisation approach that not only is infrastructure-free, but also avoids accumulation error by taking advantage of (1) the widespread ubiquity of mobile devices with cameras and (2) the availability of 3D building models for most modern buildings. Localisation is performed by matching image sequences captured by a camera, with a 3D model of the building in a model-based visual tracking framework. Comprehensive evaluation of the approach with a photo-realistic synthetic dataset shows the robustness of the localisation approach under challenging conditions. Additionally, the approach is tested and evaluated on real data captured by a smartphone. The results of the experiments indicate that a localisation accuracy better than 10 cm can be achieved by using this approach. Since localisation errors do not accumulate the proposed approach is suitable for indoor localisation tasks for long periods of time and augmented reality applications, without requiring any local infrastructure. A MATLAB implementation can be found on 〈a href="https://github.com/debaditya-unimelb/BIM-Tracker" target="_blank"〉https://github.com/debaditya-unimelb/BIM-Tracker〈/a〉.〈/p〉〈/div〉 〈/div〉

Description: 〈p〉Publication date: June 2019〈/p〉 〈p〉〈b〉Source:〈/b〉 ISPRS Journal of Photogrammetry and Remote Sensing, Volume 152〈/p〉 〈p〉Author(s): Qing Wang, Hua Sun, Ruopu Li, Guangxing Wang〈/p〉 〈div xml:lang="en"〉 〈h5〉Abstract〈/h5〉 〈div〉〈p〉Traditional parametric methods for classification of land use and land cover (LULC) types using remote sensing imagery assume a global distribution model and fail to consider local variation of categorical variables. Differently, non-parametric methods do not make any statistical assumptions but are typically sensitive to the sample sizes of training sample data that usually require a high cost to collect in the field. Geostatistical classifiers, such as indicator kriging and simulation, are local variability-based methods that exhibit great potential for image-based classification of LULC types. However, variogram models required are highly sensitive to the spatial configuration of training samples as well as sample size given a study area. Moreover, when a large number of spectral variables are considered into kriging systems, modeling the variograms and cross-variograms would be problematic. To circumvent these issues, this study extended the geostatistical methods from a 2-dimensional geographic space to a 〈em〉m〈/em〉-dimensional image feature space to derive feature-space indicator variograms (FSIVs). Moreover, a novel stochastic simulation classification algorithm, Feature-Space Indicator Simulation (FSIS), was proposed and examined for classification of LULC types in Duolun County located in Inner Mongolia and in Huang-Feng-Qiao (HFQ) forest farm, Hunan of China. In Duolun, six LULC types were involved and in HFQ a complicated forest landscape consisting of nine forest types plus water, built-up area, and agricultural/bare soil, was classified. The classification results of FSIS were compared with another feature-space geostatistical classifier – feature-space indicator kriging (FSIK), a traditional parametric method – maximum likelihood (ML), a widely used nonparametric method – support vector machine (SVM), and a recently popular method – random forest (RF). The results showed that compared with ML, SVM and RF, in both study areas FSIS statistically significantly increased the accuracy of the classifications by 10.0–29.9% for percentage correct and 19.0–47.6% for Kappa statistic. Compared with FSIK, FSIS also improved the classification accuracy but the accuracy increases were relatively smaller with the percentages correct of 3.5% and 7.6% and the Kappa values of 4.6% and 8.6% for Duolun and HFQ, respectively. Moreover, FSIS led to the spatial uncertainties of the classification estimates as the quality measure of the estimates. In addition, the results also demonstrated that FSIVs were sensitive to the within-class heterogeneity but not very much to the size of training samples. Overall, FSIS exhibited the greater potential to improve the classification accuracy of LULC and forest types using remote sensing image.〈/p〉〈/div〉 〈/div〉

Description: 〈p〉Publication date: June 2019〈/p〉 〈p〉〈b〉Source:〈/b〉 ISPRS Journal of Photogrammetry and Remote Sensing, Volume 152〈/p〉 〈p〉Author(s): Xin Huang, Ying Wang〈/p〉 〈div xml:lang="en"〉 〈h5〉Abstract〈/h5〉 〈div〉〈p〉The Urban heat island (UHI) effect is an increasingly serious problem in urban areas. Information on the driving forces of intra-urban temperature variation is crucial for ameliorating the urban thermal environment. Although prior studies have suggested that urban morphology (e.g., landscape pattern, land-use type) can significantly affect land surface temperature (LST), few studies have explored the comprehensive effect of 2D and 3D urban morphology on LST in different urban functional zones (UFZs), especially at a fine scale. Therefore, in this research, we investigated the relationship between 2D/3D urban morphology and summer daytime LST in Wuhan, a representative megacity in Central China, which is known for its extremely hot weather in summer, by adopting high-resolution remote sensing data and geographical information data. The “urban morphology” in this study consists of 2D urban morphological parameters, 3D urban morphological parameters, and UFZs. Our results show that: (1) The LST is significantly related to 2D and 3D urban morphological parameters, and the scattered distribution of buildings with high rise can facilitate the mitigation of LST. Although sky view factor (SVF) is an important measure of 3D urban geometry, its influence on LST is complicated and context-dependent. (2) Trees are the most influential factor in reducing LST, and the cooling efficiency mainly depends on their proportions. The fragmented and irregular distribution of grass/shrubs also plays a significant role in alleviating LST. (3) With respect to UFZs, the residential zone is the largest heat source, whereas the highest LST appears in commercial and industrial zones. (4) Results of the multivariate regression and variation partitioning indicate that the relative importance of 2D and 3D urban morphological parameters on LST varies among different UFZs and 2D morphology outperforms 3D morphology in LST modulation. The results are generally consistent in spring, summer and autumn. These findings can provide insights for urban planners and designers on how to mitigate the surface UHI (SUHI) effect via rational landscape design and urban management during summer daytime.〈/p〉〈/div〉 〈/div〉

Description: 〈p〉Publication date: May 2019〈/p〉 〈p〉〈b〉Source:〈/b〉 ISPRS Journal of Photogrammetry and Remote Sensing, Volume 151〈/p〉 〈p〉Author(s): Tuomas Yrttimaa, Ninni Saarinen, Ville Luoma, Topi Tanhuanpää, Ville Kankare, Xinlian Liang, Juha Hyyppä, Markus Holopainen, Mikko Vastaranta〈/p〉 〈div xml:lang="en"〉 〈h5〉Abstract〈/h5〉 〈div〉〈p〉Dead wood is a key forest structural component for maintaining biodiversity and storing carbon. Despite its important role in a forest ecosystem, quantifying dead wood alongside standing trees has often neglected when investigating the feasibility of terrestrial laser scanning (TLS) in forest inventories. The objective of this study was therefore to develop an automatic method for detecting and characterizing downed dead wood with a diameter exceeding 5 cm using multi-scan TLS data. The developed four-stage algorithm included (1) RANSAC-cylinder filtering, (2) point cloud rasterization, (3) raster image segmentation, and (4) dead wood trunk positioning. For each detected trunk, geometry-related quality attributes such as dimensions and volume were automatically determined from the point cloud. For method development and validation, reference data were collected from 20 sample plots representing diverse southern boreal forest conditions. Using the developed method, the downed dead wood trunks were detected with an overall completeness of 33% and correctness of 76%. Up to 92% of the downed dead wood volume were detected at plot level with mean value of 68%. We were able to improve the detection accuracy of individual trunks with visual interpretation of the point cloud, in which case the overall completeness was increased to 72% with mean proportion of detected dead wood volume of 83%. Downed dead wood volume was automatically estimated with an RMSE of 15.0 m〈sup〉3〈/sup〉/ha (59.3%), which was reduced to 6.4 m〈sup〉3〈/sup〉/ha (25.3%) as visual interpretation was utilized to aid the trunk detection. The reliability of TLS-based dead wood mapping was found to increase as the dimensions of dead wood trunks increased. Dense vegetation caused occlusion and reduced the trunk detection accuracy. Therefore, when collecting the data, attention must be paid to the point cloud quality. Nevertheless, the results of this study strengthen the feasibility of TLS-based approaches in mapping biodiversity indicators by demonstrating an improved performance in quantifying ecologically most valuable downed dead wood in diverse forest conditions.〈/p〉〈/div〉 〈/div〉 〈div xml:lang="en"〉 〈h5〉Graphical abstract〈/h5〉 〈div〉〈p〉〈figure〉〈img src="https://ars.els-cdn.com/content/image/1-s2.0-S0924271619300760-ga1.jpg" width="383" alt="Graphical abstract for this article" title=""〉〈/figure〉〈/p〉〈/div〉 〈/div〉

Description: 〈p〉Publication date: May 2019〈/p〉 〈p〉〈b〉Source:〈/b〉 ISPRS Journal of Photogrammetry and Remote Sensing, Volume 151〈/p〉 〈p〉Author(s): Hoang Long Nguyen, David Belton, Petra Helmholz〈/p〉 〈div xml:lang="en"〉 〈h5〉Abstract〈/h5〉 〈div〉〈p〉Plane detection and segmentation is one of the most crucial tasks in point cloud processing. The output from this process can be used as input for further processing steps, such as modelling, registration and calibration. However, the sparseness and heterogeneity of Mobile Laser Scanning (MLS) point clouds may lead to problems for existing planar surfaces detection and segmentation methods. This paper proposes a new method that can be applicable to detect and segment planar features in sparse and heterogeneous MLS point clouds. This method utilises the scan profile patterns and the planarity values between different neighbouring scan profiles to detect and segment planar surfaces from MLS point clouds. The proposed method is compared to the three most state-of-the-art segmentation methods (e.g. RANSAC, a robust segmentation method based on robust statistics and diagnostic principal component analysis – RDCPA as well as the plane detection method based on line arrangement). Three datasets are used for the validation of the results. The results show that our proposed method outperforms the existing methods in detecting and segmenting planar surfaces in sparse and heterogeneous MLS point clouds. In some instances, the state-of-the-art methods produce incorrect segmentation results for façade details which have a similar orientation, such as for windows and doors within a façade. While RDCPA produces up to 50% of outliers depending on the neighbourhood threshold, another method could not detect such features at all. When dealing with small features such as a target, some algorithms (including RANSAC) were unable to perform segmentation. However, the propose algorithm was demonstrated to detect all planes in the test data sets correctly. The paper shows that these mis-segmentations in other algorithms may lead to significant errors in the registration process of between 1.047 and 1.614 degrees in the angular parameters, whereas the propose method had only resulted in 0.462 degree angular bias. Furthermore, it is not sensitive to the required method parameters as well as the point density of the point clouds.〈/p〉〈/div〉 〈/div〉

Description: 〈p〉Publication date: May 2019〈/p〉 〈p〉〈b〉Source:〈/b〉 ISPRS Journal of Photogrammetry and Remote Sensing, Volume 151〈/p〉 〈p〉Author(s): Jianfeng Huang, Xinchang Zhang, Qinchuan Xin, Ying Sun, Pengcheng Zhang〈/p〉 〈div xml:lang="en"〉 〈h5〉Abstract〈/h5〉 〈div〉〈p〉Automated extraction of buildings from remotely sensed data is important for a wide range of applications but challenging due to difficulties in extracting semantic features from complex scenes like urban areas. The recently developed fully convolutional neural networks (FCNs) have shown to perform well on urban object extraction because of the outstanding feature learning and end-to-end pixel labeling abilities. The commonly used feature fusion or skip-connection refine modules of FCNs often overlook the problem of feature selection and could reduce the learning efficiency of the networks. In this paper, we develop an end-to-end trainable gated residual refinement network (GRRNet) that fuses high-resolution aerial images and LiDAR point clouds for building extraction. The modified residual learning network is applied as the encoder part of GRRNet to learn multi-level features from the fusion data and a gated feature labeling (GFL) unit is introduced to reduce unnecessary feature transmission and refine classification results. The proposed model - GRRNet is tested in a publicly available dataset with urban and suburban scenes. Comparison results illustrated that GRRNet has competitive building extraction performance in comparison with other approaches. The source code of the developed GRRNet is made publicly available for studies.〈/p〉〈/div〉 〈/div〉

Description: 〈p〉Publication date: May 2019〈/p〉 〈p〉〈b〉Source:〈/b〉 ISPRS Journal of Photogrammetry and Remote Sensing, Volume 151〈/p〉 〈p〉Author(s): Yusheng Xu, Richard Boerner, Wei Yao, Ludwig Hoegner, Uwe Stilla〈/p〉 〈div xml:lang="en"〉 〈h5〉Abstract〈/h5〉 〈div〉〈p〉To ensure complete coverage when measuring a large-scale urban area, pairwise registration between point clouds acquired via terrestrial laser scanning or stereo image matching is usually necessary when there is insufficient georeferencing information from additional GNSS and INS sensors. In this paper, we propose a semi-automatic and target-less method for coarse registration of point clouds using geometric constraints of voxel-based 4-plane congruent sets (V4PCS). The planar patches are firstly extracted from voxelized point clouds. Then, the transformation invariant, 4-plane congruent sets are constructed from extracted planar surfaces in each point cloud. Initial transformation parameters between point clouds are estimated via corresponding congruent sets having the highest registration scores in the RANSAC process. Finally, a closed-form solution is performed to achieve optimized transformation parameters by finding all corresponding planar patches using the initial transformation parameters. Experimental results reveal that our proposed method can be effective for registering point clouds acquired from various scenes. A success rate of better than 80% was achieved, with average rotation errors of about 0.5 degrees and average translation errors less than approximately 0.6 m. In addition, our proposed method is more efficient than other baseline methods when using the same hardware and software configuration conditions.〈/p〉〈/div〉 〈/div〉

Description: 〈p〉Publication date: May 2019〈/p〉 〈p〉〈b〉Source:〈/b〉 ISPRS Journal of Photogrammetry and Remote Sensing, Volume 151〈/p〉 〈p〉Author(s): Sugandh Chauhan, Roshanak Darvishzadeh, Mirco Boschetti, Monica Pepe, Andrew Nelson〈/p〉 〈div xml:lang="en"〉 〈h5〉Abstract〈/h5〉 〈div〉〈p〉Rapid and quantitative assessment of crop lodging is important for understanding the causes of the phenomena, improving crop management, making better production and supporting loss estimates in general. Accurate information on the location and timing of crop lodging is valuable for farmers, agronomists, insurance loss adjusters, and policymakers. Lodging studies can be performed to assess the impact of lodging events or to model the risk of occurrence, both of which rely on information that can be acquired by field observations, from meteorological data and from remote sensing (RS). While studies applying RS data to assess crop lodging dates back three decades, there has been no comprehensive review of the status, potential, current approaches, and challenges in this domain. In this position paper, we review the trends in field/lab-based and RS-based studies for crop lodging assessment and discuss the strengths and weaknesses of current approaches. Theoretical background on crop lodging is presented, and the scope of RS in assessing plant characteristics associated with lodging is reviewed and discussed. The review focuses on RS-based studies, grouping them according to the platform deployed (i.e., ground-based, airborne and spaceborne), with an emphasis on analyzing the pros and cons of the technology. Finally, the challenges, research gaps, perspectives for future research, and an outlook on new sensors and platforms are presented to provide state-of-the-art and future scenarios of RS in lodging assessment. Our review reveals that the use of RS techniques in crop lodging assessment is still in an experimental stage. However, there is increasing interest within the RS scientific community (based on the increased rate of publications over time) to investigate its use for crop lodging detection and risk mapping. The existing satellite-based lodging assessment studies are very few, and the operational application of the current approaches over large spatial extents seems to be the biggest challenge. We identify opportunities for future studies that can develop quantitative models for estimating lodging severity and mapping lodging risk using RS data.〈/p〉〈/div〉 〈/div〉

Description: 〈p〉Publication date: September 2019〈/p〉 〈p〉〈b〉Source:〈/b〉 ISPRS Journal of Photogrammetry and Remote Sensing, Volume 155〈/p〉 〈p〉Author(s): Anh Vu Vo, Debra F. Laefer, Aljosa Smolic, S.M. Iman Zolanvari〈/p〉 〈div xml:lang="en"〉 〈h5〉Abstract〈/h5〉 〈div〉〈p〉This paper presents a complete data processing pipeline for improved urban solar potential estimation by applying solar irradiation estimation directly to individual aerial laser scanning (ALS) points in a distributed computing environment. Solar potential is often measured by solar irradiation – the amount of the Sun’s radiant energy received at the Earth’s surface over a period of time. To overcome previous limits of solar radiation estimations based on either two-and-a-half-dimensional raster models or overly simplistic, manually-generated, geometric models, an alternative approach is proposed using dense, urban aerial laser scanning data to enable the incorporation of the true, complex, and heterogeneous elements common in most urban areas. The approach introduces a direct, per-point analysis to fully exploit all details provided by the input point cloud data. To address the resulting computational demands required by the thousands of calculations needed per point for a full-year analysis, a distributed data processing strategy is employed that introduces an atypical data partition strategy. The scalability and performance of the approach are demonstrated on a 1.4-billion-point dataset covering more than 2 km〈sup〉2〈/sup〉 of Dublin, Ireland. The reliability and realism of the simulation results are rigorously confirmed with (1) an aerial image collected concurrently with the laser scanning, (2) a terrestrial image acquired from an online source, and (3) a four-day, direct solar radiation collection experiment.〈/p〉〈/div〉 〈/div〉

Description: 〈p〉Publication date: September 2019〈/p〉 〈p〉〈b〉Source:〈/b〉 ISPRS Journal of Photogrammetry and Remote Sensing, Volume 155〈/p〉 〈p〉Author(s): Alfonso Fernandez-Manso, Carmen Quintano, Dar A. Roberts〈/p〉 〈div xml:lang="en"〉 〈h5〉Abstract〈/h5〉 〈div〉〈p〉All ecosystems and in particular ecosystems in Mediterranean climates are affected by fires. Knowledge of the drivers that most influence burn severity patterns as well an accurate map of post-fire effects are key tools for forest managers in order to plan an adequate post-fire response. Remote sensing data are becoming an indispensable instrument to reach both objectives. This work explores the relative influence of pre-fire vegetation structure and topography on burn severity compared to the impact of post-fire damage level, and evaluates the utility of the Maximum Entropy (MaxEnt) classifier trained with post-fire EO-1 Hyperion data and pre-fire LiDAR to model three levels of burn severity at high accuracy. We analyzed a large fire in central-eastern Spain, which occurred on 16–19 June 2016 in a maquis shrubland and 〈em〉Pinus halepensis〈/em〉 forested area. Post-fire hyperspectral Hyperion data were unmixed using Multiple Endmember Spectral Mixture Analysis (MESMA) and five fraction images were generated: char, green vegetation (GV), non-photosynthetic vegetation, soil (NPVS) and shade. Metrics associated with vegetation structure were calculated from pre-fire LiDAR. Post-fire MESMA char fraction image, pre-fire structural metrics and topographic variables acted as inputs to MaxEnt, which built a model and generated as output a suitability surface for each burn severity level. The percentage of contribution of the different biophysical variables to the MaxEnt model depended on the burn severity level (LiDAR-derived metrics had a greater contribution at the low burn severity level), but MaxEnt identified the char fraction image as the highest contributor to the model for all three burn severity levels. The present study demonstrates the validity of MaxEnt as one-class classifier to model burn severity accurately in Mediterranean countries, when trained with post-fire hyperspectral Hyperion data and pre-fire LiDAR.〈/p〉〈/div〉 〈/div〉

Description: 〈p〉Publication date: September 2019〈/p〉 〈p〉〈b〉Source:〈/b〉 ISPRS Journal of Photogrammetry and Remote Sensing, Volume 155〈/p〉 〈p〉Author(s): Hasan Asy’ari Arief, Ulf Geir Indahl, Geir-Harald Strand, Håvard Tveite〈/p〉 〈div xml:lang="en"〉 〈h5〉Abstract〈/h5〉 〈div〉〈p〉Advances in techniques for automated classification of point cloud data introduce great opportunities for many new and existing applications. However, with a limited number of labelled points, automated classification by a machine learning model is prone to overfitting and poor generalization. The present paper addresses this problem by inducing controlled noise (on a trained model) generated by invoking conditional random field similarity penalties using nearby features. The method is called Atrous XCRF and works by forcing a trained model to respect the similarity penalties provided by unlabeled data. In a benchmark study carried out using the ISPRS 3D labeling dataset, our technique achieves 85.0% in term of overall accuracy, and 71.1% in term of F1 score. The result is on par with the current best model for the benchmark dataset and has the highest value in term of F1 score. Additionally, transfer learning using the Bergen 2018 dataset, without model retraining, was also performed. Even though our proposal provides a consistent 3% improvement in term of accuracy, more work still needs to be done to alleviate the generalization problem on the domain adaptation and the transfer learning field.〈/p〉〈/div〉 〈/div〉

Description: 〈p〉Publication date: October 2019〈/p〉 〈p〉〈b〉Source:〈/b〉 ISPRS Journal of Photogrammetry and Remote Sensing, Volume 156〈/p〉 〈p〉Author(s): Ethan Heck, Kirsten M. de Beurs, Braden C. Owsley, Geoffrey M. Henebry〈/p〉 〈div xml:lang="en"〉 〈h5〉Abstract〈/h5〉 〈div〉〈p〉The latest collection (C6) of MODIS data provides several algorithmic improvements and calibration adjustments that correct for sensor degradation, theoretically making the C6 MODIS products more accurate compared to previous collections. C6 adjustments also introduce several improvements in the vegetation index (VI) retrieval algorithms. With these improvements, we expect only minor differences between data from Terra and Aqua, but significantly different results between C5 and C6. In this paper, we investigate three different MODIS products to determine the extent that improvements made to C6 influence the overall trend results for time series between 2001 and 2017. We focus on these three products specifically, both to allow for a comparison of vegetation index products—NDVI and EVI from MOD13C1, and NDVI and EVI calculated based on surface reflectance from MCD43C4—and also to gain an understanding of the improvements on an entirely different product from the same sensor, namely Land Surface Temperature (LST) from MOD11C2. For the MCD43C4 dataset, we find that 17.9% and 16.4% of EVI and NDVI pixels, respectively, display trend discordance between C5 and C6. For the MOD13C1 vegetation indices, we found comparable rates of trend discordance between C5 and C6: 18.5% and 17.4% for the EVI and NDVI pixels, respectively. For both products the greatest changes between C5 and C6 are an overall increase in pixels exhibiting a significant greening trend and an overall decline in pixels exhibiting a significant browning trend. Moreover, the largest differences between C5 and C6 for the NDVI and EVI data appear in cropland areas and in regions with relatively little human influence. In the Land Surface Temperature product (MOD11C2), the discordance between C5 and C6 is much lower: only 3.2% of day and 5.0% of night LST trends exhibited discordance between C5 and C6. We analyze the complementary results of vegetation index and land surface temperature trends and demonstrate that combining the results from different products observed at different portions of the electromagnetic spectrum—but linked through the biogeophysical processes of surface energy balance—allows us to portray change with more confidence than when relying on vegetation index data alone.〈/p〉〈/div〉 〈/div〉

Description: 〈p〉Publication date: October 2019〈/p〉 〈p〉〈b〉Source:〈/b〉 ISPRS Journal of Photogrammetry and Remote Sensing, Volume 156〈/p〉 〈p〉Author(s): Bin Chen, Yufang Jin, Patrick Brown〈/p〉 〈div xml:lang="en"〉 〈h5〉Abstract〈/h5〉 〈div〉〈p〉Floral phenology, the timing and intensity of flowering, is intimately tied to the reproduction of terrestrial ecosystem and highly sensitive to climate change. However, observational records of flowering are very sparse, limiting our understanding of spatiotemporal dynamics of floral phenology from local to regional scales. Satellite remote sensing provides unique opportunities to monitor flowers through space and time in a cost-effective way. Here we developed an enhanced bloom index (EBI), based on the multispectral remotely sensed data, to quantify flowering status over almond (〈em〉Prunus dulcis〈/em〉) orchards in Central Valley of California. Our test studies with unmanned aerial vehicle (UAV) multispectral imagery at 2.6–5.2 cm demonstrated that the EBI enhanced the signals of flowers and reduced the background noise from soil and green vegetation, and agreed well with the bloom coverage derived from supervised classification, with a R〈sup〉2〈/sup〉 of 0.72. Experimental tests with multi-scale remote sensing observations from CERES aerial (0.2 m), PlanetScope (3 m), Sentinel-2 (10 m), and Landsat (30 m) satellite imagery further showed the robustness of the EBI in capturing the flower information. We found that the relatively dense time series of PlanetScope and Sentinel-2 imagery were able to capture the bloom dynamics of almond orchards. Satellite derived EBI is expected to track the bloom information and thus improve our understanding and prediction of flower and pollination response to weather and ultimately the yield.〈/p〉〈/div〉 〈/div〉 〈div xml:lang="en"〉 〈h5〉Graphical abstract〈/h5〉 〈div〉〈p〉〈figure〉〈img src="https://ars.els-cdn.com/content/image/1-s2.0-S092427161930190X-ga1.jpg" width="439" alt="Graphical abstract for this article" title=""〉〈/figure〉〈/p〉〈/div〉 〈/div〉

Description: 〈p〉Publication date: October 2019〈/p〉 〈p〉〈b〉Source:〈/b〉 ISPRS Journal of Photogrammetry and Remote Sensing, Volume 156〈/p〉 〈p〉Author(s): Davoud Ashourloo, Hamid Salehi Shahrabi, Mohsen Azadbakht, Hossein Aghighi, Hamed Nematollahi, Abbas Alimohammadi, Ali Akbar Matkan〈/p〉 〈div xml:lang="en"〉 〈h5〉Abstract〈/h5〉 〈div〉〈p〉Different techniques utilized for mapping various crops are mainly based on using training dataset. But, due to difficulties of access to a well-represented training data, development of automatic methods for detection of crops is an important need which has not been considered as it deserves. Therefore, main objective of present study was to propose a new automatic method for canola (〈em〉Brassica napus〈/em〉 L〈em〉.〈/em〉) mapping based on Sentinel 2 satellite time series data. Time series data of three study sites in Iran (Moghan, Gorgan, Qazvin) and one site in USA: (Oklahoma), were used. Then, spectral reflectance values of canola in various spectral bands were compared with those of the other crops during the growing season. NDVI, Red and Green spectral bands were successfully applied for automatic identification of canola flowering date using the threshold values. Examination of the fisher function indicated that multiplication of the near-infrared (NIR) band by the sum of red and green bands during the flowering date is an efficient index to differentiate canola from the other crops. The Kappa and overall accuracy (OA) for the four study sites were more than 0.75 and 88%, respectively. Results of this research demonstrated the potential of the proposed approach for canola mapping using time series of remotely sensed data.〈/p〉〈/div〉 〈/div〉

Description: 〈p〉Publication date: September 2019〈/p〉 〈p〉〈b〉Source:〈/b〉 ISPRS Journal of Photogrammetry and Remote Sensing, Volume 155〈/p〉 〈p〉Author(s): Yanhua Xie, Tyler J. Lark, Jesslyn F. Brown, Holly K. Gibbs〈/p〉 〈div xml:lang="en"〉 〈h5〉Abstract〈/h5〉 〈div〉〈p〉Accurate and timely information on the distribution of irrigated croplands is crucial to research on agriculture, water availability, land use, and climate change. While agricultural land use has been well characterized, less attention has been paid specifically to croplands that are irrigated, in part due to the difficulty in mapping and distinguishing irrigation in satellite imagery. In this study, we developed a semi-automatic training approach to rapidly map irrigated croplands across the conterminous United States (CONUS) at 30 m resolution using Google Earth Engine. To resolve the issue of lacking nationwide training data, we generated two intermediate irrigation maps by segmenting Landsat-derived annual maximum greenness and enhanced vegetation index using county-level thresholds calibrated from an existing coarse resolution irrigation map. The resulting intermediate maps were then spatially filtered to provide a training data pool for most areas except for the upper midwestern states where we visually collected samples. We then used random samples extracted from the training pool along with remote sensing-derived features and climate variables to train ecoregion-stratified random forest classifiers for pixel-level classification. For ecoregions with a large training pool, the procedure of sample extraction, classifier training, and classification was conducted 10 times to obtain stable classification results. The resulting 2012 Landsat-based irrigation dataset (LANID) identified 23.3 million hectares of irrigated croplands in CONUS. A quantitative assessment of LANID showed superior accuracy to currently available maps, with a mean Kappa value of 0.88 (0.75–0.99), overall accuracy of 94% (87.5–99%), and producer’s and user’s accuracy of the irrigation class of 97.3% and 90.5%, respectively, at the aquifer level. Evaluation of feature importance indicated that Landsat-derived features played the primary role in classification in relatively arid regions while climate variables were important in the more humid eastern states. This methodology has the potential to produce annual irrigation maps for CONUS and provide insights into the field-level spatial and temporal aspects of irrigation.〈/p〉〈/div〉 〈/div〉

Description: 〈p〉Publication date: August 2019〈/p〉 〈p〉〈b〉Source:〈/b〉 ISPRS Journal of Photogrammetry and Remote Sensing, Volume 154〈/p〉 〈p〉Author(s): Bashar Elnashef, Sagi Filin〈/p〉 〈div xml:lang="en"〉 〈h5〉Abstract〈/h5〉 〈div〉〈p〉Accuracy, detail, and limited time on site make photogrammetry a valuable means for underwater mapping. Imaging in such domains is subjected however to distortions which are caused by refraction of the incoming rays. As the literature shows, these distortions are depth-dependent, non-linear, and alter the standard single viewpoint geometry. To handle their effect, we derive in this paper a refraction-invariant representation and show that despite the pronounced distortions, such a model is attainable. We also show that its contribution is not only theoretical, as it also allows to estimate the pose parameters linearly and at a significantly improved accuracy. The paper then extends the model to calibrate the underwater-related system parameters and, again, demonstrates the ability to yield a linear model, to simplify the settings and requirements for calibration procedures, and most importantly to improve the accuracy of the system parameters by an order of magnitude or more. Experiments also show enhanced accuracy and stability of the model in the presence of high-level of noise. Thus, the paper provides an in-depth look into the geometrical modeling of underwater images and at the same time offers practical enhancement of the accuracies and requirements to reach them.〈/p〉〈/div〉 〈/div〉

Description: 〈p〉Publication date: August 2019〈/p〉 〈p〉〈b〉Source:〈/b〉 ISPRS Journal of Photogrammetry and Remote Sensing, Volume 154〈/p〉 〈p〉Author(s): Rongjun Qin〈/p〉 〈div xml:lang="en"〉 〈h5〉Abstract〈/h5〉 〈div〉〈p〉The geometric analysis and data acquisition of satellite photogrammetric images are often regarded as a direct extension of traditional aerial photogrammetry, with the only difference being the sensor model (linear array vs. central perspective). The intersection angle (or base-height ratio) between two images is seen as the most important metadata of stereo pairs, which directly relates to the base-high ratio and texture distortion in the parallax direction, thus both affecting the horizontal and vertical accuracy. State-of-the-art DIM algorithms were reported to work best for narrow baseline stereos (small intersection angle), e.g. Semi-Global Matching empirically takes 15–25° as “good” intersection angles. However, our experiments found that the intersection angle is not the only determining factor, as the same DIM algorithm applied to stereo pairs of the same area with similar and good intersection angle may produce point clouds with dramatically different accuracy (demonstrated in the graphical abstract). 〈strong〉〈em〉This raises a very practical and often asked question: what factors constitute a good satellite stereo pair for DIM algorithms?〈/em〉〈/strong〉 In this paper, we provide a comprehensive analysis on this matter by performing stereo matching using the very typical and widely-used Semi-Global Matching (SGM) with a Census cost over 1000 satellite stereo pairs of the same region with different meta-parameters including their intersection, off-nadir, sun elevation & azimuth angles, completeness and time differences, thus to offer a thorough answer to this question. 〈strong〉〈em〉Our conclusion has specifically outlined an important yet often ignored factor – the Sun-angle difference to be one decisive in determining good stereo pair.〈/em〉〈/strong〉 Based on the analytical results, we propose a simple idea by training a support vector machine model for predicting potential stereo matching quality (i.e. potential level of accuracy and completeness given a stereo pair). Experiments have shown that the model is well-suited and generalized for multi-stereo 3D reconstruction, evidenced by a comparative analysis against three other strategies: (1) pair selection based on an example patch where partial ground-truth data is available for computing a priori ranking (2) based on intersection angles and (3) based on a recent algorithm using intersection angle, off-nadir angle and time intervals. This work will potentially provide a valuable reference to researchers working on multi-view satellite image reconstruction, as well as for practitioners minimizing costs for high-quality large-scale mapping. The trained model is made available to the academic community upon request.〈/p〉〈/div〉 〈/div〉 〈div xml:lang="en"〉 〈h5〉Graphical abstract〈/h5〉 〈div〉 〈p〉Digital surface models generated using Semi-global matching (SGM, with Census as the cost metric) from different stereo pairs with a sample of different meta-information. We demonstrate that the meta-information for stereo pair is critical. In addition to well-known factors such as intersection angles, the sun angle plays a significant role in dense matching results (as highlighted in red). It is shown in the last images of the two rows that the stereo pair being “in-track” or “cross-track” does not significantly impact the results much as long as their meta-parameters are similar.〈/p〉 〈p〉〈figure〉〈img src="https://ars.els-cdn.com/content/image/1-s2.0-S0924271619301467-ga1.jpg" width="247" alt="Graphical abstract for this article" title=""〉〈/figure〉〈/p〉 〈/div〉 〈/div〉

Description: 〈p〉Publication date: August 2019〈/p〉 〈p〉〈b〉Source:〈/b〉 ISPRS Journal of Photogrammetry and Remote Sensing, Volume 154〈/p〉 〈p〉Author(s): Chunping Qiu, Lichao Mou, Michael Schmitt, Xiao Xiang Zhu〈/p〉 〈div xml:lang="en"〉 〈h5〉Abstract〈/h5〉 〈div〉〈p〉The local climate zone (LCZ) scheme was originally proposed to provide an interdisciplinary taxonomy for urban heat island (UHI) studies. In recent years, the scheme has also become a starting point for the development of higher-level products, as the LCZ classes can help provide a generalized understanding of urban structures and land uses. LCZ mapping can therefore theoretically aid in fostering a better understanding of spatio-temporal dynamics of cities on a global scale. However, reliable LCZ maps are not yet available globally. As a first step toward automatic LCZ mapping, this work focuses on LCZ-derived land cover classification, using multi-seasonal Sentinel-2 images. We propose a recurrent residual network (Re-ResNet) architecture that is capable of learning a joint spectral-spatial-temporal feature representation within a unitized framework. To this end, a residual convolutional neural network (ResNet) and a recurrent neural network (RNN) are combined into one end-to-end architecture. The ResNet is able to learn rich spectral-spatial feature representations from single-seasonal imagery, while the RNN can effectively analyze temporal dependencies of multi-seasonal imagery. Cross validations were carried out on a diverse dataset covering seven distinct European cities, and a quantitative analysis of the experimental results revealed that the combined use of the multi-temporal information and Re-ResNet results in an improvement of approximately 7 percent points in overall accuracy. The proposed framework has the potential to produce consistent-quality urban land cover and LCZ maps on a large scale, to support scientific progress in fields such as urban geography and urban climatology.〈/p〉〈/div〉 〈/div〉

Description: 〈p〉Publication date: August 2019〈/p〉 〈p〉〈b〉Source:〈/b〉 ISPRS Journal of Photogrammetry and Remote Sensing, Volume 154〈/p〉 〈p〉Author(s): Shishi Liu, Hang Su, Guofeng Cao, Shanqin Wang, Qingfeng Guan〈/p〉 〈div xml:lang="en"〉 〈h5〉Abstract〈/h5〉 〈div〉〈p〉Annual analyses of land cover dynamics in urban areas provide a thorough understanding of the urbanization effects on environment and valuable information for the improvement of urban growth modeling. However, most current studies focus on major land cover changes, such as urbanization and vegetation loss. The most feasible way to evaluate the complex interactions among different land cover types is the post-classification change detection, but the temporal inconsistency in the time series of land cover maps impedes the high-frequency and long-term analyses. This study proposed a spatio-temporal land cover filter (STLCF) to remove the illogical land cover change events in the time series of land cover maps, and analyzed the annual land cover dynamics in urban areas. The knowledge of illogical land cover change events was ‘learned’ from the land cover maps through the spatio-temporal transition probability matrix, instead of experts’ knowledge. The illogical change was modified with the land cover of the maximum probability calculated from the naïve Bayesian equation. The STLCF was tested in Wuhan, a typical densely urbanized Chinese city. The annual land cover maps from 2000 to 2013 were derived from multi-date Landsat images using the Decision Tree (DT) classifier. Results showed that the STLCF improved the mean overall accuracy of annual change detection by about 6%. Additionally, the amount of land cover trajectories with unrealistically frequent changes was significantly decreased. During the study period, 7.86% of the pixels experienced one land cover change, and about 0.57% of the pixels experienced land cover changes more than once. The annual analyses demonstrated the non-linear increasing trend in urbanization as well as the corresponding trend in vegetation loss in the study area. We also found the conversion from built-up areas to vegetation near rivers and lakes and in the reserves and rural areas, mainly caused by the restoration of built-up areas to the park or green belt/wedges along rivers and new roads in the metropolitan areas, and to the cropland and woods in the rural areas. Results of this study showed the importance of the spatio-temporal consistency check with knowledge derived from land cover maps of the study area, which facilitates the annual analyses of major and subtle land cover dynamics in urban areas.〈/p〉〈/div〉 〈/div〉 〈div xml:lang="en"〉 〈h5〉Graphical abstract〈/h5〉 〈div〉〈p〉〈figure〉〈img src="https://ars.els-cdn.com/content/image/1-s2.0-S0924271619301479-ga1.jpg" width="291" alt="Graphical abstract for this article" title=""〉〈/figure〉〈/p〉〈/div〉 〈/div〉

Description: 〈p〉Publication date: August 2019〈/p〉 〈p〉〈b〉Source:〈/b〉 ISPRS Journal of Photogrammetry and Remote Sensing, Volume 154〈/p〉 〈p〉Author(s): Jibo Yue, Jia Tian, Qingjiu Tian, Kaijian Xu, Nianxu Xu〈/p〉 〈div xml:lang="en"〉 〈h5〉Abstract〈/h5〉 〈div〉〈p〉Soil moisture (SM) controls the exchange of water and heat energy between the land surface and the atmosphere through evaporation and plant transpiration, and timely and accurate estimates of soil moisture are crucial for many studies. Although remote sensing provides many algorithms to obtain bare-soil moisture on large scales (e.g., Soil Moisture and Ocean Salinity (SMOS) and Soil Moisture Active and Passive (SMAP) products), these algorithms are restricted to studies with low ground resolution. In the present study, we propose and evaluate three normalized shortwave-infrared (SWIR) difference bare soil moisture indices [NSDSI1 = (B〈sub〉SWIR1〈/sub〉-B〈sub〉SWIR2〈/sub〉)/B〈sub〉SWIR1〈/sub〉, NSDSI2 = (B〈sub〉SWIR1〈/sub〉 − B〈sub〉SWIR2〈/sub〉)/B〈sub〉SWIR2〈/sub〉, NSDSI3 = (B〈sub〉SWIR1〈/sub〉 − B〈sub〉SWIR2〈/sub〉)/(B〈sub〉SWIR1〈/sub〉 + B〈sub〉SWIR2〈/sub〉), where, SWIR1 = 1550–1750 nm, SWIR2 = 2100–2300 nm] to estimate the bare-soil moisture content and, based on the water absorption difference between shortwave-infrared bands, use them to map bare-soil moisture with high ground resolution Sentinel-2 MSI images. The three proposed bare-soil moisture indices are obtained by using the different water absorption in shortwave-infrared bands. Four traditional hyperspectral-based bare-soil moisture indices (such as the water index SOIL, or WISOIL, the normalized soil moisture index, or NSMI, etc.) were used as benchmark. The results show that (i) the differences in water absorption between shortwave-infrared bands is linear in soil moisture content; (ii) the combined use of two shortwave-infrared bands from four soils provides more accurate bare-soil moisture estimates than do single shortwave-infrared (SWIR) bands, and (iii) our proposed bare-soil moisture indices can be applied on broadband remote-sensing images (such as Landsat, Sentinel-2 MSI). The bare-soil moisture estimates obtained by using the proposed bare-soil moisture indices may help to obtain bare-soil moisture maps with high ground resolution.〈/p〉〈/div〉 〈/div〉