SAR Time-domain Imaging Algorithm Design And System Development Based On Embedded GPU

In recent years,there has demand for SAR imaging technology in many fields.With the development of warfare,the target's anti-missile defense capabilities continue to increase,and different types of interference continue to increase.The missile-borne radar seeker is facing a more complex battlefield environment.This requires that the radar seeker can quickly image the target of interest in order to obtain fine information of the target in real time and accurately identify it.In the missile-borne platform,in order to achieve high maneuverability,on the one hand,the SAR imaging algorithm usually needs to work in the front-side view mode.The radar echo signal will have a serious two-dimensional coupling between the range and the azimuth.The frequency domain SAR imaging algorithm is used for imaging.The effect is poor,so it is necessary to use a time-domain imaging method with higher imaging quality;on the other hand,the SAR imaging algorithm must complete the processing in a short time to enable the radar seeker to quickly identify and locate the target.Therefore,it is of great significance to use a higher-performance GPU platform for SAR time-domain imaging processing.This article mainly elaborates the design and system development process of SAR time-domain imaging algorithm based on embedded GPU from three aspects: the principle and improvement method of time-domain imaging algorithm,the implementation of embedded GPU of imaging algorithm,and the design and development of digital simulation test system.First,the accelerated decomposition back projection algorithm based on spectrum fusion is introduced.Aiming at the problem of low AFBP time domain imaging speed,the polynomial fitting method is used to accelerate the calculation of the slope distance and the angle of view,which effectively improves the imaging speed.For the problem of platform motion error,the motion compensation algorithm based on inertial navigation data and the phase gradient autofocus algorithm based on echo data are studied.Finally,the effectiveness of the improved AFBP algorithm are verified through experiments.Then,in order to further improve the time of SAR imaging,the Jetson AGX Xavier,a more powerful embedded GPU platform,was introduced,and several GPU efficient programming and optimization methods were proposed,which were compared with the TMS320C6678 chip in an all-round way,and based on The embedded GPU platform implements and optimizes the improved AFBP imaging algorithm.Compared with the implementation using FPGA+multi-chip DSP platform,the implementation speed-up ratio of the embedded GPU platform is as high as 5.4 times,indicating that this article is based on the SAR time domain of the embedded GPU.The imaging implementation has great advantages in real-time computing.Finally,in order to be able to verify the SAR time-domain imaging function based on the embedded GPU platform,simulation system was developed.The system cooperated with the embedded GPU signal processing platform to form a semi-physical simulation test system.Run the test and the results show that the time-domain imaging results of the embedded GPU platform described in this article are basically consistent with the MATLAB imaging results,and meet the real-time requirements,and can run stably and smoothly.