Abstract:

This paper proposed a novel segmentation method which integrates statistical distribution, geometric shape features and polarimetric decomposition features for high resolution polarimetric synthetic aperture radar (SAR) data. This method is based on the fractal network evolution algorithm (FNEA) that integrates K distribution statistics and Pauli decomposition features. Specifically, statistical heterogeneity of objects is defined by the maximum log likelihood function based on K distribution. Polarimetric decomposition heterogeneity of objects is calculated through the weighted sum of standard deviation of Pauli decomposition features. A total heterogeneity of objects is defined by the weighted sum of statistical heterogeneity, polarimetric decomposition heterogeneity and shape heterogeneity. Then, the multi-feature segmentation procedure for high resolution polarimetric SAR data is constructed. The effectiveness of the integrated multi-feature segmentation we develope is demonstrated by simulated data and L band ESAR polarimetric data.