GPU-based Airborne SAR Fine Imaging Application System

At present,airborne synthetic aperture radar(SAR)imaging has reached sub-meter resolution,and the amount of echo data and signal calculation is huge in fine processing.Traditionally methods relying on single/multi-core CPU processing cannot meet the realtime imaging,and it is urgent to introduce a high-performance computing platform to achieve fast high-resolution imaging.The GPU is made up of thousands of small,efficient hardware cores that excel at processing large numbers of concurrent and lightweight threads and excel in performing massively parallel experiments.The SAR imaging algorithm usually exhibits signal-level concurrency,which provides a feasible condition for the application of the GPU to accelerate the algorithm.The SAR imaging processing technology on the GPU came into being.Aiming at this hot topic,it's studied that the imaging radar parallel acceleration technology and radar interface display technology on CPU+GPU.Combine the two,the GPU-based SAR fine processing imaging system is designed.The main work of the thesis is:First,it is introduced that the line frequency modulation(CS)algorithm and the CS fine imaging algorithm combined with motion compensation.The functional segment of inertial compensation,Doppler center estimation,distance migration correction,Doppler frequency estimation,and full aperture imaging are studied one by one.And its signal level concurrency is analyzed.Second,The CPU+GPU heterogeneous parallel development technology is studied,and the GPU acceleration model matched with the SAR imaging algorithm is designed.The CUDA C programming experiment is completed by simulation.Then focus on the Doppler frequency estimation function code segment,combined with Nsight software to test its performance,and give the optimization strategy of delay hidden and Time/spectrum shift conversion according to the specific hardware resources..Third,the Qt technology is used,a GPU-based fine imaging software was designed,and the CUDA parallel program of SAR imaging was loaded to display the imaging results.The functions cover button control,parameter display,GPU calculation status display,and image display.The software satisfies the requirements of human-computer interaction to control GPU calculation mapping.GPU-based airborne SAR fine imaging application system is a combination of highresolution SAR imaging algorithm,GPU high-performance parallel computing,Qt interface API.The system fully schedules the GPU hardware resources,sets up reasonable thread configuration,and performs the parallel computing of SAR imaging processing,which reduces the complexity of the algorithm,improves the execution efficiency,and improves the data processing speed and hardware resource utilization as a whole.Compared with the single-core 3.4G 8-core CPU i7 processor running MATLAB algorithm,using the Kepler architecture K40 C graphics card CPU + GPU heterogeneous platform computing,the acceleration effect is more than 60 times.