A real-time ISAR imaging structure based on GPU and CPU heterogeneous parallel processing

To achieve the real-time processing requirement of high resolution ISAR imaging, this paper introduces a GPU and CPU heterogeneous parallel processing structure. In this processing structure, four GPUs are utilized to process parallel computing intensive tasks of ISAR imaging such as pulse compression, range alignment, Doppler phase correction and azimuth FFT, while the CPU is utilized to do the logic operation in the processing and to calculate target parameters that ISAR imaging required. The detailed implementation of the range-Doppler (RD) algorithm in this structure is presented in this paper. Simulation results verify the validity and efficiency of the processing structure.