Publication Date:
2019
Description:
This research aims to assess the capabilities of Very High Spatial Resolution (VHSR)hyperspectral satellite data in order to discriminate urban tree diversity. Four dimension reductionmethods and two classifiers are tested, using two learning methods and applied with four in situsample datasets. An airborne HySpex image (408 bands/2 m) was acquired in July 2015 from whichprototypal spaceborne hyperspectral images (named HYPXIM) at 4m and 8m and a multispectralSentinel2 image at 10 m have been simulated for the purpose of this study. A comparison is madeusing these methods and datasets. The influence of dimension reduction methods is assessed onhyperspectral (HySpex and HYPXIM) and Sentinel2 datasets. The influence of conventional classifiers(Support Vector Machine –SVM– and Random Forest –RF–) and learning methods is evaluated on allimage datasets (reduced and non-reduced hyperspectral and Sentinel2 datasets). Results show thatHYPXIM 4 m and HySpex 2 m reduced by Minimum Noise Fraction (MNF) provide the greatestclassification of 14 species using the SVM with an overall accuracy of 78.4% (±1.5) and a kappa indexof agreement of 0.7. More generally, the learning methods have a stronger influence than classifiers,or even than dimensional reduction methods, on urban tree diversity classification. PrototypalHYPXIM images appear to present a great compromise (192 spectral bands/4 m resolution) for urbanvegetation applications compared to HySpex or Sentinel2 images.
Electronic ISSN:
2072-4292
Topics:
Architecture, Civil Engineering, Surveying
,
Geography
Permalink