Research On SAR Image Feature Extraction And Classification Method

SAR (Synthetic Aperture Radar) is a kind of active microwave sensor. It occupies a veryimportant position in imaging radar and has already been widely applied to national economy,national defense and scientific research, etc. At present, it has become a research hotspot toclassify and interpreter a ground scene based on SAR image data. In this paper, in-depthresearches have been done on SAR image feature extraction and SAR image classification.The main research contents are shown as follows:1. Aiming at the extraction of texture feature of SAR image, the traditional GLCM (GreyLevel Co-occurrence Matrix) algorithm and some improved algorithms are analyzed andcompared. In order to solve the problem of time complexity of these algorithms, the parallelrapid extraction algorithm of GLCM texture feature has been studied in this paper by takingadvantage of the parallel performance of FPGA (Field Programmable Gate Array). Thealgorithm takes advantage of the high speed parallel computing of FPGA to extract texturefeatures of SAR images quickly and efficiently. Especially for large-scaled SAR images, theefficiency of feature extraction is higher. The efficiency and accuracy of FPGA-based GLCMalgorithm have been verified through experiments.2. On the basis of analyzing and comparing supervised classification, unsupervisedclassification and semi-supervised classification algorithms, by combining NJW(Ng-Jordan-Weiss) spectral clustering algorithm and PSO (Particle Swarm Optimization)algorithm, a semi-supervised particle swarm spectral clustering algorithm has been proposedin this paper. Different algorithms are compared to realize precise SAR image classification. Ithas been proved through experiments that the proposed algorithm is superior to the traditionalNJW algorithm.