A SAR moving targets imaging algorithm design and implementation based on CUDA

In the application of SAR moving targets imaging, typical algorithms are generally complex, requiring powerful floating-point computing ability and a lot of time. Starting with analysis of the error of conventional still targets imaging algorithms, this paper designs a moving targets imaging algorithm by error compensation and motion parameters estimation. This algorithm is simple but efficient, and easily parallelized as well. Therefore, we implement the algorithm based on the widely used Compute Unified Device Architecture (CUDA) platform and test this implementation with data acquired from real SAR system. Tests prove the effectiveness of this algorithm and the implementation on CUDA shows very high computational efficiency. For every testing moving target, the implementation is able to finish the imaging processing within one second.