An Improved Algorithm to Delineate Urban Targets with Model-Based Decomposition of PolSAR Data

Abstract

In model-based decomposition algorithms using polarimetric synthetic aperture radar (PolSAR) data, urban targets are typically identified based on the existence of strong double-bounced scattering. However, urban targets with large azimuth orientation angles (AOAs) produce strong volumetric scattering that appears similar to scattering characteristics from tree canopies. Due to scattering ambiguity, urban targets can be classified into the vegetation category if the same classification scheme of the model-based PolSAR decomposition algorithms is followed. To resolve the ambiguity and to reduce the misclassification eventually, we introduced a correlation coefficient that characterized scattering mechanisms of urban targets with variable AOAs. Then, an existing volumetric scattering model was modified, and a PolSAR decomposition algorithm developed. The validity and effectiveness of the algorithm were examined using four PolSAR datasets. The algorithm was valid and effective to delineate urban targets with a wide range of AOAs, and applicable to a broad range of ground targets from urban areas, and from upland and flooded forest stands.