Classification of multi-look polarimetric SAR data based on complex Wishart distribution

Abstract

An optimal feature classification scheme is developed for multilook polarimetric SAR (synthetic aperture radar) imagery based on a multivariate complex Wishart distribution. The purpose is to identify various ground covers, such as forest, vegetation, city block, ocean, and sea ice type. Multilook polarimetric SAR data can be represented either in Stoke's matrix form or in complex covariance matrix form. The latter has a complex Wishart distribution. A simple but effective classifier is then developed using the complete information of the complex covariance. This algorithm is further extended to classification using multifrequency polarimetric data. A procedure for assessing the classification errors is also developed using a Monte Carlo simulation. The effectiveness of this algorithm is demonstrated with NASA/JPL (Jet Propulsion Laboratory) P-, L-, and C-band polarimetric SAR data.<>