Shoreline classification using dual-polarized TerraSAR-X images

Abstract:
In this paper we present a new method for shoreline classification based on dual-polarized X-band SAR data. The crucial point of the study is that a partial Kennaugh matrix has been established as a suitable decomposition for co- and cross-polarized images in order to highlight polarimetric features. An additional improvement of radiometric stability is provided by a new multi-scale multi-looking approach. The resulting images can easily be introduced to a supervised maximum likelihood land cover classification. The resulting classes are compared to ground truth data collected during the summer months along transects. This work proves the suitability of multipolarized SAR data for land cover classification in sub-Arctic regions. It is a collaboration of the German Remote Sensing Data Center and Environment Canada.