In this paper, a new method for SAR image classification based on decision tree and Wishart classifier was proposed. Firstly, feature parameters were extracted from the fully polarimetric SAR data. Next, the extracted feature parameters were used as input of the decision tree. Then the classification result was classified as the initial class of the Wishart classifier. Compared with the H-α decomposition, this method not only retains the physical scattering mechanism of classification results, but also can effectively classify sandy land features. The effectiveness of this method is demonstrated using C-band fully polarimetric SAR data of Hunshandake Sandy Land, acquired by the RADARSAT-2 sensor.
|