Synthetic aperture radar feature selection for dual polarized ScanSAR data

Synthetic aperture radar (SAR) ScanSAR data has advantages on oceanographic remote sensing applications regarding its large coverage and sufficient resolution. However terrestrial downlink bandwidth is limited and therefore up to dual polarized (e.g., HH and HV) ScanSAR data can be achieved today's spaceborne systems (e.g., RadarSAT-2). In this study grey level co-occurrence matrix was employed to extract SAR features for both HH and HV channels. Additionally some of band math products such as HH/HV and HH-HV were used as candidate SAR features. Selection of optimum SAR features is crucial and application dependent. In this study, selection strategies based on SAR data assimilation was introduced and relation of conventional separability criterions on SAR data assimilation and pattern recognition were discussed.