Polarimetric SAR image classification based on target decomposition theorem and complex Wishart distribution

Polarimetric SAR data have been frequently utilized to classify terrain types. This paper compares two terrain classification approaches: 1) classification based on the target decomposition theorem, and 2) classification based on the multivariate complex Wishart distribution. Target decomposition theorems were developed recently for extracting geophysical information of scattering media. The averaged Mueller matrix or its equivalent was decomposed into components for a thorough study. Parameters were introduced to characterize the physical aspect of the scattering process and they are used for unsupervised classification. From a different point of view, the statistical property of the polarimetric covariance matrix has been modeled with a complex Wishart distribution. Supervised and unsupervised classification based on this distribution reported good results. The authors present results of supervised classifications using both methods. Comparisons are made on the accuracy of these two classification schemes. Possible reasons for the cause of errors are discussed. The NASA/JPL San Francisco data is chosen for illustration