Unsupervised classification based on the logarithmic circular polarization ratio parameter for hybrid polarimetric SAR

In this study, we investigated the unsupervised terrain classification for hybrid polarimetric (HP) SAR by using the circular polarization ratio (CPR) parameter alone. According to the theoretical deduction based on scattering matrices of ideal polarimetric scattering mechanisms (PSMs), CPR is suggested to be processed with the logarithmic function in order to have a balanced span between the ideal PSMs' CPR values, which in turn can improve the identification of real PSMs. Utilizing one simulated HP dataset, the performance of the proposed logarithmic CPR parameter is first compared with that of classical m and χ parameters, and then assessed with real PSM classes identified by the H-α classification algorithm. Finally, a simple classification scheme of terrains is proposed and validated.