Airborne PolarSAR Classification Based On The Analysis Of The Features Scattering Properties

As a new pattern SAR technique, the PolSAR technique represents more comprehensive information for surface features. With the advantage of weather resistant, this technique has been a hot research topic in recent years. There are more and more applications in relevant fields. One of the most important fields is the PolSAR classification. After several years’ research, the experts from home and abroad have come up with some milestone theories and algorithms. However, there are still some problems in PolSAR classification techniques. This thesis aims at two of these problems, contributing our views on the classification using PolSAR data, especially the airborne PolSAR data.Firstly, as the comprehension of the polarization scattering properties is not sufficient, our thesis analyze various classical polarization scattering parameters under different kinds of surface features systematically, and lay particular emphasis on their change pattern with the variation of the incident angle parameter.Secondly, we aim at the problem that existent airborne PolSAR classification methods reckon without the clear variation of incident angle. We propose a classification method with the compensation of the incident angle for the first time. We employ several airborne PolSAR data from different platforms to validate the effectiveness and the reliability of our method, and analyze the feasibility under different wave bands.In addition, rely on the Western China Mapping Project; we utilize two classical classifiers (Wishart classifier and SVM classifier) to qualitatively compare the classification results under the condition of whether correct the incident angle or not. On account of our method is not applicable in mountainous area, we offer an idea of Decision tree classification base on the Freeman-Durden decomposition. The results show that this idea has some reference value.