Unsupervised Classification of Polarimetric Sar Data Using Image Clustering and H/a/α Decomposition

This paper proposed a method of unsupervised classification of polarimetric SAR data based on image clustering and H/A/α decomposition. Fully polarimetric L band data collected by ALOS PALSAR system was used in this paper. The study site locates in Tianjin, China. The area is characterized by rural residence, bare soil, water body and pond. An unsupervised classification method is proposed for fully polarimetric SAR data. The relation between physical structure and polarimetric signal properties is studied explicitly using polarimetric decomposition. Consequently, a method based on clustering and iteration was introduced and was shown to yield radar-derived maps with a high level of agreement with existing TM images. * Corresponding author LiangWenjing Wang. [email protected]