SAR Target Discrimination And Recognition Algorithm Based On SAR-SIFT Feature

Synthetic aperture radar(SAR)can detect target under all weather,all time conditions,from long distance,and acquire a large number of high-resolution SAR images.With the application of SAR imaging technology in military and civil fields,SAR image target discrimination and recognition technology have become the focus of research.In practice,because the scene around the target and the structure of the target are complex,the traditional target discrimination algorithms can not accurately identify the target from scene,and the false alarm rate is high.In the process of recognition,the performance of the traditional target recognition algorithms need to be further improved due to the presence of multiplicative speckle noise and a large number of target variants in the SAR image.In view of above problems,this thesis studies the SAR target discrimination and recognition algorithms based on the SAR-SIFT feature.At the stage of discrimination,the acquired discrimination feature improves the target discrimination rate,and reduces the clutter false alarm as well.At the stage of recognition,the feature vector and the coordinates of the feature points are effectively combined to improve the recognition rate of the target variants.The main contents of this thesis are summarized as follows:1.This thesis introduces the background,significance and development of SAR image target discrimination and target recognition,and summarizes the main work of the thesis.2.This thesis introduces the traditional SIFT algorithm and the SAR-SIFT algorithm for SAR images.In the experiments,the property of the edge detection by the difference gradient and the edge detection by the ROEWA are analysed,and the performance of different keypoint detectors is analyzed.Finally,the advantages of SAR-SIFT feature extraction algorithm in SAR image are verified by different combinations of keypoint detectors and feature descriptors.3.This thesis studies SAR target discrimination algorithm based on SAR-SIFT feature.Firstly,the SPM algorithm and Sc SPM algorithm are introduced.Then,the two-parameter CFAR detection algorithm and the two-parameter CFAR detection algorithm based on super-pixel are introduced to extract the target chips and clutter chips from the SAR image;the SAR-SIFT algorithm and Sc SPM algorithm are combined to extract the discrimination feature.Finally,the linear SVM classifier is used to discriminate the chips extracted by different detection algorithms,and improved results are obtained.4.This thesis studies the SAR target recognition algorithm based on SAR-SIFT feature.Firstly,the target image from the MSTAR data set is processed to complete the registration by translation,in order to move the target to the center of the image.Then,the registered image is processed to complete the binary segmentation,and the target region in the image is segmented from the clutter background.Then,according to the scattering characteristics of the vehicle target in the SAR image,the strong scattering points on the target are extracted based on the binary segmentation image.At the same time,the SAR-Harris corner detection algorithm is used to extract the corner points on the target based on the binary segmentation image.Then the strong scattering points and the corner points on the target are taken as the feature points.Finally,the featrue vectors are generated according to the SAR-SIFT feature extraction algorithm,and then the LTS-HD distance of the feature vectors and the LTS-HD distance of the feature points are fused to obtain better target recognition results.