Radar high range resolution profile (HRRP) contains target structure information, and RATR based on HRRP has become a hotly researched topic. As high dimension of HRRP, the method of independent component analysis (ICA) is proposed to extract some features of HRRP, which is decomposed in the field of "base signals", and the coefficients are extracted to form vectors as the target's feature vectors. The Back-Propagation Network (BP) is employed as the classifier. As the classifier may classify the non- depository target sample, the output of BP network is encoded and the refuse-recognition threshold is proposed in this paper, a new classifier is born when the refuse-recognition threshold is adopted. The classifier can refuse the non- depository target samples and put the depository target samples into its class correctly. The simulation of some targets echoes is done by finite difference time domain (FDTD). The experiment results show that the method proposed performs well.