Multi-core DSP-based Acceleration of SAR BP Algorithm Imaging

Abstract

The back projection (BP) imaging algorithm is a time-domain synthetic aperture radar (SAR) imaging algorithm. The BP algorithm does not perform approximation processing during imaging and has high imaging accuracy. However, the BP algorithm needs sub-image imaging at each azimuth moment, which has a high computational complexity and limits the application in scenarios with high imaging rate requirements. Therefore, in this paper we proposed a parallel acceleration scheme of the BP algorithm based on the FT-M3000 digital signal processor (DSP) to parallelize the computation of sub-images at each azimuth moment in sub-aperture imaging, and realize parallel imaging of four sub-aperture data. Experimental results show that the total imaging time of the four sub-aperture images is 4.12 seconds, compared with the imaging time of the four sub-aperture images using a 3.3GHz CPU for 263.65 seconds, and the imaging speed of this scheme is increased by 64 times.