Fuzzy clustering and interpretation of fully polarimetric SAR data

Abstract

Classification of Earth terrain components using fully polarimetric synthetic aperture radar (SAR) data sets is an important application of radar remote sensing. For the operational application some demands, besides the accuracy requirements, must be fulfilled. In order to make the handling of the classification easy for users, the algorithms have to be data set independent and the handling must be possible without a priori knowledge. The ultimate aim is an unsupervised algorithm which is suitable for automation. In this treatment we propose an approach applying an unsupervised automatic clustering of the H/A//spl alpha///spl lambda//sub 1/ space. From the resulting clusters rules are derived for a fuzzy rule based classification. The resulting clusters can then be assigned to the desired object classes by the user. The approach enables us to combine the wide range of information contained in polarimetric SAR data with the robust and still flexible strategy of fuzzy rule based classifiers and with a high degree of automation. The effectiveness of this approach is demonstrated using fully-polarimetric L-band airborne SAR data acquired with the E-SAR system of the DLR at the well know test site of Oberpfaffenhofen, Germany.