Improved subspace method for fully polarimetric SAR image classification

This paper proposes an improved subspace method (ISM) for fully polarimetric synthetic aperture radar (PolSAR) image classification, which is a combination of the leaning subspace method (LSM), the averaged learning subspace method (ALSM), and the multiple similarity method (MSM). The fully polarimetric Radarsat-2 image for the Yellow River Delta of northern Shandong Province is selected to evaluate the recognition accuracy. The supervised Wishart method is also performed for comparison. Experimental results validate the proposed method yielded better classification results. Therefore, the ISM is a feasible method for fully polarimetric SAR image classification.