Different targets in space have different physical and geometry characteristics. RCS(Radar Cross-Section)is one of the most important characteristics of targets, so it is a very important method to classify targets based on RCS data. LPC(Linear Predictive Coding)is used to process RCS data in order to obtain LPC coefficients of them. Then the LPC coefficients are used as feature vectors for targets classification and recognition. In this way, the classification rate can be up to 80%~90% when classifying two targets. And the classification rate can be up to 50%~80% when classifying three targets. This method can effectively improve the radar automatic classification capability so as to track targets semi-automatically and automatically.