Non-Parameter Correlation Analysis in Polarimetric Signature and its Application to Change Detection in Polarimetric SAR

Gonradsen et al. developed a change detection method with multilook fully polarimetric Synthetic Aperture Radar (SAR) data. The method is based on complex Wishart distribution supposition limits the applicable scope of test statistic, especially when only polarimetric diversity (the texture and speckle statistics depend on polarization) and not radiometric diversity. On the other hand, the polarization signature has the capability of representation of the polarimetric diversity. A new test statistic for equality of two polarization signature is proposed in this paper. Without statistical distribution hypothesis, it can give the associated asymptotic probability measure as the Conradsen´s method just by non-parameter correlation analysis. Some experiments have been done in this paper with ALOS-PALSAR data by two different algorithms. It has been shown that the new method is effective.