LAUSR

This work aims at addressing the overestimation of volume scattering in model-based decomposition of hybrid-polarimetric (hybrid-pol) Synthetic Aperture Radar (SAR) data. Usually, in the hybrid-pol decomposition techniques reported in the literature, complete depolarized power has been considered as volume scattering power contribution. Assigning all the depolarized power to the volume scattering mechanism results in an overestimation of the volume scattering contribution. Further, this results in the scattering mechanism ambiguity and in turn, diminishes the classification accuracy for areas dominated by dihedral scattering and surface scattering. In order to address this overestimation problem, in this work, three popular volume scattering models are introduced into the Hybrid-pol Three-Component Scattering Model (HTM) framework. The three proposed algorithms are compared against each other and with the classical hybrid-pol methods, namely $m-\delta $ , $m-\chi $ , and modified $m-\chi $ decomposition. In addition, the performance of the original HTM algorithm is improved by replacing the pseudo quad-pol reconstruction process used therein with an entropy-based reconstruction approach available in the literature. This algorithm is referred to as “Modified HTM” in this work. One of the proposed algorithms is inspired by the Generalized Volume Scattering Model. It is found to provide a better average classification accuracy as compared to both, the classical HTM and the modified HTM algorithm. A relative assessment of all the algorithms has also been included based on ship detection. All the algorithms are tested using the hybrid-pol L-band dataset synthesized from ALOS-PALSAR 2 mission.