Support Vector Machine And Its Applied Research In Radar Target Recognition

This thesis mainly studies the problem of Automatic Target Recognition (ATR) based on Support Vector Machine(SVM) Our research mainly concentrates on the method of target's feature extraction and the design of classifier based on Support Vector Machine.At the beginning, the thesis makes a brief introduction to the development of radar target recognition technique, feature extraction and the design of classifier are the pivotal points. Based on the general models of the radar target scatter and the predigested scatter models under the special conditions, we use Relax arithmetic to extract the scatter's position information from radar HRRP as the feature vector. DCT is introduced to the HRRP spectral estimation. Then ,we use RELAX parameters as feature vector, SVM as classifier to form a high resolution radar target recognition system. In the SAR target recognition system .For the SAR images on the same target at different aspect angles are not very similar to each other. We use SAR release data as feature, design three classifiers as the perceptron ,the optimal hyperplane and the SVM classifier to form the SAR target recognition system.The computer simulation using MSTR and the ISAR data shows that the identification accuracy is very good. The average identifying rate exceeds 97.93%and 91.55 %under the condition that the SVM contain complete information of radar target.