Research On The Target Recognition In SAR Images

Synthetic aperture radar(SAR) is a system of high-altitude visual which has characters as all-weather, long distance, strong penetrable ability and high resolution. SAR target recognition has become a hotspot. SAR-ATR is an important aspect of military reconnaissance, its process contains detection, discrimination and classification, in which target classification of SAR images is one of the key techniques. Thus the study of target classification in SAR images is valued and significant. The paper explores the thoughts and methods of target classification at the aspects of template matching and Support Vector Machine(SVM).1. Using the simulation algorithm to imaging the different types of aircraft models under different azimuth in order to build a simulation template library. We use the real SAR images of an airport in Shanxi and simulation template library as a database in the experiment. Meanwhile, using Lee filter, Kuan filter, Sigma filter, Frost filter and GMAP algorithm to filter the speckle in the SAR images, then, using eleven methods to extract the features of the database.2. In order to improve the performance of classifiers in the SAIP system, a two-stage template-based classifier is studied. The classifier screens several templates which have the higher matching rate by using the minimum mean square error(MSE) algorithm. Then, the classifier uses the Burg-MSE algorithm to enhances images’ resolution and find the template which has the highest matching rate.3. In order to use the SVM classifier to solve the two-class classification problem and the multi-class classification problem, a classifier that combines the basic SVM classifier with the binary tree is studied, then, using the classifier to recognition the SAR targets. The experiment indicates that the SVM classifier has a higher recognition accuracy.