Frequency Decomposition POLSAR Image Unsupervised Classification Method Based Study

Polarimetric SAR image classification is an important research direction of SAR image processing. Compared with the SAR image, Polarimetric SAR images contain not only the amplitude and phase information of ground clutters, but also a wealth of polarization information. The effect of polarimetric SAR image classification largely depends on the accurate description of the ground clutter scattering mechanism. In this paper, we first study the basic theoretical of polarimetric SAR images, then analyze the existing polarimetric SAR image classification methods and problems, then we proposed a classification method, which is based on the time-frequency decomposition. This method can express the scattering features more accurately and completely. This method can also extract the coherent interference information from the mining of one image. The classification result of this method is better than the traditional methods, especially in the discrimination of the woodland and manmade building areas. The main work of this paper is as follows:1) Introduced a time-frequency decomposition method into the extracting of the non-stationary and coherent scattering features in the scene. We used the azimuth time-frequency decomposition method to analyze the scattering characteristics of non-stationary target in the scene. Then we extracted the scattering characteristics of non-stationary. We used the range direction time-frequency decomposition method to analyze the scattering characteristics of coherent scatters. After that we extracted the scattering characteristics of coherent scattering.2) Proposed an unsupervised classification method based on time-frequency decomposition and Wishart classifier. This method is more accurate in description of the feature properties in the scene, thereby improved the classification accuracy of polarimetric SAR images. This method played a good role on the classification of forest and building area.3) Quantitative evaluation method was introduced to study the classification results of the polarimetric SAR unsupervised classification methods. Experiments show that this method can achieve a better classification result, especially in the classification of woodland and building.