SAR AUTOMATIC TARGET RECOGNITION BASED ON KPCA CRITERION
Han Ping Wu Renbiao Wang Zhaohua Wang Yunhong (Institute of Electronic Information Eng., Tianjin University, Tianjin 300072, China) (Inst. of Comm. and Signal Proc., Civil Aviation Univ. of China, Tianjin 300300, China) (Nat. Key Lab of Pattern Recognition, Inst. of Automation, CAS, Beijing 100080, China)
In this paper, SAR ATR (Synthetic Aperture Radar Automatic Target Recognition) approach based on KPCA (Kernel Principal Component Analysis) is proposed. KPCA first maps the input data into some feature space using kernel functions and then performs linear PCA on the mapped data. It takes the principal components in nonlinear space as sample features, then SVM classifier is used to classify targets. Experimental results with MSTAR SAR data sets provided by the US DARPA/AFRL (Defense Advanced Research Projects Agency/Air Force Research Laboratory) show a better performance of classification and generalization.