A SAR target detection model based on CFAR and KPCA is presented in this paper. This Detection is divided into a pre-screening and discrimination process. Within the large-scale and low-resolution SAR imagery, pre-screening adopts classic CFAR techniques, while the discrimination process adopts kernel principal component analysis to separate the target from clutter. Experimental results show that the detection performance of our algorithm appears to be superior to the classic CFAR methodology. The combination of both pre-screening and prior knowledge of targets can effectively enhance detection rate and inhibit false alarm at the same time.