A polarimetric interferometric SAR image-based land cover supervised classfication method

High-resolution X-band polarimetric interferometric synthetic aperture radar (PolInSAR) images often contain strong speckle noise, which can be an obstacle for applications like land cover classification. To overcome this problem, we apply the Nonlocal filter to the images first. After that, polarimetric and interferometric features are extracted and then used by the support vector machine (SVM) and the AdaBoost classifier for classification. For demonstrating the effectiveness of the presented method, we test it on an PolInSAR image of Weinan collected by N-SAR. This area contains 10 cover types and there is enough ground truth for training and validation. Classification result shows that the AdaBoost classifier achieves good performance for various cover types, and that interferometric information can improve the accuracy.