Abstarct: In today world with confinements and limitations surveillance systems has an important rule in immunization. Military organizations and centers needs the best and the most carefully surveillance systems for importance of their missions. Uses of automatic systems in these organizations are growing widely. One of the researches in surveillance systems is Automatic Target Recognition (ATR). The goal of an ATR system is obtain information from enemy vehicles rapidly and more accurately and also with minimum risks for human agent. In this thesis we propose an accurate and powerful method for recognizing military vehicles in Synthetic Aperture Radar (SAR) images. SAR sensors can work in nights and with variations in weather condition. This thesis is organized in two main parts. First in image processing phase that include use of wavelet transform and morphology operations for noise reduction and target detection. Second part is pattern recognition phase that include feature extraction and classification operations. In this thesis we extract 8 geometrical feature from target. For target classification we use some of supervised classifiers. Finally we show the effect of using ensemble learning algorithms on improving classification result. We implemented our method with data from MSTAR public dataset. The result is compared with the results of three articles and demonstrated the accuracy and reliability of our method.