Unsupervised classification for compact polarimetric SAR data using m-δ decomposition, SPAN and the Wishart classifier

This paper studies the polarimetric decomposition and unsupervised classification of compact polarimetric data. Compact polarimetric data have been synthesized from AirSAR L-band fully polarimetric data by matrix transformations. Polarimetric decomposition for compact polarimetric data have been introduced and analyzed. Decomposition result of relevant fully polarimetric data has been used for comparison. An unsupervised classification method is proposed for compact polarimetric data using polarimtric decomposition. Degree of polarization m and relative H-V phase δ combined with the total backscattering power SPAN have been used to initialize the compact polarimetric data into 15 classes. Then the Wishart classifier is applied. The result of this algorithm is illustrated and compared with the classification result of full polarimetric data.