Classification for POLSAR Images Based on Surface Scattering Similarity

Aiming at the deficiency of the classification based on H-Alpha polarimetric decomposition theorem, we study the classification for POLSAR images based on scattering similarity. Firstly, a new parameter is proposed to measure the scattering similarity between targets. As this parameter contains all scattering similarities between the scattering mechanisms of a distributed target and canonical mechanism, it describes accurately the degree of average scattering similarity between a distributed target and a canonical target. As an application, we analyze the feasibility of the scattering similarity parameter corresponding to the sphere scattering in the extraction of average scattering theoretically. Furthermore, we also discuss the classification scheme with the scattering similarity corresponding to the sphere scattering and the polarimetric scattering entropy and give the boundary of the effective classification region theoretically. With the AIRSAR data of San Francisco, the veracity of scattering similarity in measuring the degree of average scattering similarity and the effective of the scattering similarity corresponding to the sphere scattering are validated.