Performance of a polarimetric SAR calibration algorithm

Results of an investigation of the performance of a polarimetric SAR (synthetic aperture radar) calibration algorithm are presented. This calibration technique assumes that the co- and cross-polarized components of azimuthally symmetric natural targets are uncorrelated, and that these targets are reciprocal. Monte Carlo simulations were performed to evaluate its performance for a variety of system and surface types, in the presence of various noise levels. Some empirical limits are obtained on leakage between channels, correlation between HH and VV backscatter, ratio of cross-pol to co-pol backscatter, system noise level, size of training sample, and amount of correlation between cross-pol and co-pol, for which good results are obtained.