ALBERT

Description: In classification of satellite images acquired over smallholder agricultural landscape with complex spectral profiles of various crop types, exploring image spatial information is important. The deep convolutional neural network (CNN), originally designed for natural image recognition in the computer vision field, can automatically explore high level spatial information and thus is promising for such tasks. This study tried to evaluate different CNN structures for classification of four smallholder agricultural landscapes in Heilongjiang, China using pan-sharpened 2 m GaoFen-1 (meaning high resolution in Chinese) satellite images. CNN with three pooling strategies: without pooling, with max pooling and with average pooling, were evaluated and compared with random forest. Two different numbers (~70,000 and ~290,000) of CNN learnable parameters were examined for each pooling strategy. The training and testing samples were systematically sampled from reference land cover maps to ensure sample distribution proportional to the reference land cover occurrence and included 60,000–400,000 pixels to ensure effective training. Testing sample classification results in the four study areas showed that the best pooling strategy was the average pooling CNN and that the CNN significantly outperformed random forest (2.4–3.3% higher overall accuracy and 0.05–0.24 higher kappa coefficient). Visual examination of CNN classification maps showed that CNN can discriminate better the spectrally similar crop types by effectively exploring spatial information. CNN was still significantly outperformed random forest using training samples that were evenly distributed among classes. Furthermore, future research to improve CNN performance was discussed.

Description: Automation is an inevitable trend in the development of tunnel shotcrete machinery. Tunnel environmental perception based on 3D LiDAR point cloud has become a research hotspot. Current researches about the detection of tunnel point clouds focus on the completed tunnel with a smooth surface. However, few people have researched the automatic detection method for steel arches installed on a complex rock surface. This paper presents a novel algorithm to extract tunnel steel arches. Firstly, we propose a refined function for calibrating the tunnel axis by minimizing the density variance of the projected point cloud. Secondly, we segment the rock surface from the tunnel point cloud by using the region-growing method with the parameters obtained by analyzing the tunnel section sequence. Finally, a Directed Edge Growing (DEG) method is proposed to detect steel arches on the rock surface in the tunnel. Our experiment in the highway tunnels under construction in Changsha (China) shows that the proposed algorithm can effectively extract the points of the edge of steel arches from 3D LiDAR point cloud of the tunnel without manual assistance. The results demonstrated that the proposed algorithm achieved 92.1% of precision, 89.1% of recall, and 90.6% of the F-score.