Terrain Classification of PolSAR Images Based on Optimal Features Combination

Terrain classification of PolSAR images is an important research direction in remote sensing image interpretation. The effective extraction of useful information from the acquired PolSAR image has become a pressing challenge. In this paper, a terrain classification method of PolSAR images based on the optimal selection of polarization features and texture features is proposed. The method first extracts 96-dimensional polarization features and texture features of PolSAR image, then selects optimal features based on feature judgment factor, feature selection factor, and texture feature contrast, screens out 10-dimensional polarization features and texture features as training samples, and finally uses SVM method to realize terrain classification. The results show that the average classification accuracy of 6 types of terrain in the Flevoland dataset reaches more than 97% by using the proposed method, and the classification efficiency of terrains can be effectively improved without affecting the classification accuracy by selecting the optimal combination of polarization and texture features compared with using all features.