ISAR (Inverse Synthetic Aperture Radar inverse synthetic aperture radar) is a kind of high resolution imaging radar, the imaging process is a heavy task of processing the high density computing and a large amount of data. GPU has dozens of times the CPU floatingpoint computation capacity and transmission bandwidth.The development of CUDA technology makes the multithreads, single instruction GPU architecture can quickly for parallel computing.This paper presents an efficient method of non coherent imaging using the parallel characteristic of GPU on the CUDA platform and realizes the parallel imaging processing.Compared with the general GPU processing method, the processing speed can achieve realtime imaging effect, and the operation rate is obviously improved.