Radar target identification by kernel principal component analysis on RCS

In this paper, the radar target identification is given by KPCA (kernel principal component analysis) on RCS (radar cross section). Theoretically, the KPCA is an improved form of PCA (principal component analysis). It first transforms data from original space to eigenspace, and PCA processing is further implemented in eigenspace. The goal is to extract much information of features and reduce noises effects. The KPCA achieves nonlinear mapping through dot products of kernel functions, but not through transfer functions. Thus one can avoid the difficulty of determining nonlinear transfer functions. In this study, the KPCA is utilized to extract features' information of angular-diversity RCS (radar cross section) data from targets and then to implement target identification. Numerical simulation shows that the proposed recognition scheme is very accurate, and can well tolerate random noises.