| Video Synthetic Aperture Radar can continuously illuminate the same area to obtain dynamic observation effects of the area.Compared with traditional SAR imaging,video SAR can achieve continuous imaging,so that more complete information of the target can be observed,and it has the characteristics of high resolution and all-weather work,and has a wide range of application scenarios.In addition to imaging,the function of ground moving target detection is still important,so it is of great significance to realize real-time processing of video SAR imaging and GMTI on UAV.In this paper,the signal processing software is designed on the TMS320C6678 multi-core DSP chip,and the signal processing software realizes the real-time processing of video SAR imaging and GMTI.The signal processing board based on VPX bus structure and C6678 processor can realize high-speed data transmission and meet the processing capacity requirements.On this basis,in order to explore the effect of other high-performance computing devices in radar signal processing,this paper also uses the Jetson TX2 GPU module to implement the main process of the video SAR imaging algorithm and the GMTI algorithm.The GPU module can also meet the real-time requirements.And has stronger processing performance.In principle,this paper first analyzes the echo signal model of video SAR in spotlight mode,and then studies the basic principle of Polar Format Algorithm(PFA),and proposes an improved PFA algorithm.The algorithm uses point multiplication linear phase instead of distance interpolation for resampling,which reduces the computational complexity.Then this paper introduces the processing method of GMTI,and mainly studies the principle of constant false alarm detection.In terms of implementation,this paper implements the algorithm on two platforms,DSP and GPU.This paper introduces the processing flow and detailed implementation method of the improved PFA algorithm using DSP,including range resampling and motion compensation,azimuth interpolation,azimuth modulation frequency estimation,azimuth compression,range geometric correction,PGA estimation,azimuth equal spacing interpolation and Steps such as ground inclination correction.In addition,this paper proposes a matrix transposition method based on the original address,which does not occupy extra space and is suitable for a variety of points,which can quickly realize matrix transposition.This paper presents the realization scheme of video SAR imaging algorithm based on GPU.The PFA algorithm from distance resampling to azimuth compression is realized by using GPU,and how to use the parallel computing characteristics of GPU to write programs is put forward,and the relationship between threads and data is given.A matrix transposition method using shared memory is proposed,which can perform matrix transposition in the original address space,and the processing speed is faster.Then,the image results and processing capabilities of the DSP and GPU are verified by using the measured data.Under the same processing flow,the speed of running the video SAR imaging algorithm on the GPU is 2.8 times that of the DSP.Both processing schemes can meet the needs of video Real-time requirements for SAR imaging.Aiming at the GMTI algorithm,this paper presents the processing flow of DSP and GPU to realize the algorithm and the realization method of each step.The processing flow of DSP and GPU in GMTI mode is the same,mainly including motion compensation,Doppler center estimation,azimuth FFT,constant false alarm detection,parameter estimation and point trace pairing.In addition,this paper presents a variety of low-level optimization methods of DSP and analyzes the optimization effects of the methods based on inline functions and software pipelining.When using GPU to implement the GMTI algorithm,a heterogeneous processing method of "CPU+GPU" is proposed.The dot trace pairing is placed on the host side(host)in the GPU module,and other steps are arranged on the device side(device)processing,so that the GPU The module exerts the maximum processing performance.The results and processing capabilities of the DSP and GPU are verified by the measured data.The speed of running the GMTI algorithm on the GPU is2.5 times that of the DSP,and it meets the real-time requirements.All data results and running times based on DSP and GPU processing meet the requirements,which verify the feasibility of the design in this paper. |
