Land cover classification from ICESat/GLAS waveform data

摘要

Lidar waveform attributes extracted from the Ice, Cloud, and land Elevation Satellite/Geoscience Laser Altimeter System (ICESat/GLAS) were used to distinguish land cover types for GLAS footprints over Jakobshavn Glacier, West Greenland. The waveform derived attributes (reflectivity and kurtosis) and self-computed waveform attributes (waveform width and waveform beginning location) were extracted as features, while true labels were generated from the LANDSAT 7 image as ground truth through gap-filling, three-band combination and supervised maximum likelihood classification. A Gaussian Process (GP) classifier was used to train classification model for classifying the land cover types into three different categories: snow, bare bedrock and sea water. Over this test site, the algorithm achieved an overall accuracy (OA) of 92.22%.