An efficient multi-receiver synthetic aperture sonar imaging algorithm for large data in heterogeneous environment

Abstract Synthetic aperture sonar is one kind of high resolution imaging sonar, which has been widely used in both military and civil applications. However, with the continuous improvement of imaging resolution, the amount of data processed per unit time is increasing, which seriously affects the application of synthetic aperture sonar. After analyzing the characteristics of each processing step of range Doppler imaging algorithm, such as range compression, fixed phase compensation, range cell migration compensation and azimuth compression, an efficient imaging algorithm in heterogeneous environment is proposed in this paper. CPU is mainly responsible for the control of the imaging process, while GPU is responsible for the intensive computation during the imaging processing process, which jointly realize the improvement of imaging efficiency. In order to eliminate the limitation of GPU memory, we proposed a blocking strategy, which can effectively improve the adaptability of the proposed imaging algorithm. We divide the key processing steps into blocks according to the data, upload the local data blocks to GPU memory in turn, and download them to the host memory after been processed, which can effectively avoid the limitation of GPU memory for large data processing. In order to verify the performance of the proposed algorithm, simulation imaging experiment is used to verify the correctness of the parallel imaging algorithm, real synthetic aperture sonar data imaging experiments are used to verify the efficiency of the proposed algorithm, and blocked parallel imaging processing with different number of pulse is simulated to verify the correctness and efficiency. The experimental results show that the real time acceleration rate of the proposed algorithm is above 14, which can meet the signal processing requirements of real time synthetic aperture sonar imaging system.