Unsupervised terrain classification preserving polarimetric scattering characteristics
Synthetic Aperture Radar (SAR) data can be difficult to interpret over glacierized terrain, especially in areas of high relief where most of the smaller glaciers of the world are found. The two primary objectives of this paper are to correct the topographically induce distortions inherent in SAR image of rugged terrain and to map glacier facies in the corrected polarimetric (PolSAR) image. Methods for terrain correction of PolSAR developed in this paper, which base on the coherency matrix data, utilize polarimetric signature to refine the location accuracy and perform the pixel size normalization on each element of the coherency matrix using the backward integration method. Moreover, azimuth-slope correction is immediately following after radiometric correction. A supervised classification is performed on the orthorectified image to map the spatial distribution of snow and ice. The visual boundary between wet snow and bare ice surface is delineated on the orthorectified image too. These results show that glacier facies and the wet snow line can be mapped reasonably well using Radarsat-2 PolSAR in a mountainous environment. * Corresponding author.