Independent component analysis of polarimetric SAR data for separating ground and vegetation components

Abstract

In this study, the use of independent component analysis (ICA) to separate the ground and vegetation scattering components from polarimetric, synthetic aperture radar (SAR) data is presented. In vegetation covered terrain, the backscattered signal is a composite of both a ground scattering component and a vegetation scattering component. The FastICA algorithm is customized and applied to both simulated and field data. The improved customized algorithm is found to work better than the standard FastICA algorithm. It is also shown that noise and depolarization limit the ability of ICA to separate the different scattering components. The technique is applied to field data from the Glen Afric region in Scotland and produces results consistent with other radar metrics obtained from that region.