Real-Time Imaging Scheme of Short-Track GB-SAR Based on GPU + OpenMP

The subaperture real-time imaging scheme proposed in this article adopts parallel architecture and processing method. This scheme is divided into three processing directions: host stream, blocking stream, and nonblocking stream, which greatly improves the parallelism of the algorithm and processing efficiency. Range compression is performed through a combination of hardware and software to reduce the burden on the software. Keystone interpolation is performed from the direction of the estimated data to complete the range cell migration correction (RCMC) and achieve the focus of the range direction. The azimuth direction is completely focused by the azimuth block, and the six kernel functions of the block part run on multiple host threads scheduled by OpenMP, which can effectively reduce the time complexity of the block part. The verification experiment based on the measured data shows that the algorithm has good focusing performance and acceleration performance. The speedup can reach 20–27, and the relative error of amplitude and phase is close to 0. With the increase in the detection range, the acceleration effect will be further improved. Finally, at the range of 1338.5 m, the speedup tends to be saturated.