A New Preprocessing Method for Polarimetric SAR Calibration

Some polarimetric calibration algorithms require reference distribution targets with reciprocity and reflectance symmetry, but usually only some natural objects in images satisfy reflection symmetry. In this Letter, a polarimetric calibration preprocessing method is proposed to extract targets with reflective symmetry in images, by combining a SVM classifier with two polarimetric features and a sample coarse extraction method, which realizes the accurate calculation of polarimetric distortion parameters. Experiments were carried out using Radarsat-2 full polarimetric data, and the results show that after preprocessing, the calculated polarimetric distortion parameters are not only stable but also improve in accuracy.