Detecting depolarizing targets with satellite data: A new geometrical perturbation filter

Target detectors using polarimetry are often focused on single (coherent) targets, since these are the ones that can be more simply characterized polarimetrically. The new proposed algorithm is aimed at the more difficult problem of partial target detection (i.e. targets with any degree of polarization). A new feature vector is defined starting from the coherency matrix, and then a perturbation method is performed. Starting from the partial target detection, a novel classification algorithm is proposed. The validation is carried out against fully polarimetric satellite data. In particular, X band TerraSAR-X and L band ALOS PALSAR are employed, providing significant agreement with the expected results and the supervised Wishart classifier.