Research On GPU-based SAR Echo Simulation Imaging And Image Quality Improvement

Synthetic Aperture Radar(SAR)is a common remote sensing system,widely used in military and civilian fields,and can detect and image long-range targets in various weather conditions.In the SAR field,there are many important technologies,such as echo signal simulation,imaging,and sidelobe suppression.Among them,SAR echo simulation is a computer simulation technology that simulates the echo signal in a real scene.The SAR system can be verified and evaluated through the simulated echo.In terms of SAR imaging,the back-projection algorithm is a typical time-domain algorithm.Since there is no approximation in the processing,it can obtain more accurate imaging than the frequencydomain algorithm,but the amount of calculation is huge.In terms of SAR image processing,we usually hope to obtain SAR images with lower sidelobe energy for subsequent image recognition and other operations.However,as the amount of data increases,the SAR technology mentioned above is difficult to perform fast and real-time processing on the traditional CPU platform,and cannot be applied in actual projects.Based on the Graphics Processing Unit(GPU)and CUDA parallel programming architecture,this thesis studies and parallel designs the concentric circle echo simulation algorithm,backprojection algorithm,and sidelobe suppression algorithm,which improves its computing efficiency.The main work of this thesis is as follows:1.Introduced the programming mode,software and hardware models and corresponding relationships under the CUDA architecture.The characteristics of CUDA storage types such as shared memory,global memory,registers,etc.are described,and the CUDA program optimization acceleration strategy is analyzed from the aspects of memory access optimization and thread setting.2.The related theories of SAR echo signal are studied,including the geometric model of SAR,chirp signal and pulse compression,and the expression of SAR echo signal is given.Based on this,the time-domain pulse coherence method and the concentric circle method are studied,and the specific simulation procedures of the two algorithms are summarized.The parallelism of the concentric circle algorithm is analyzed.Based on the CUDA architecture,it is designed in parallel,and its performance is improved by optimizing the variable storage and algorithm structure,which improves the computing efficiency of echo simulation.The simulation proves the effectiveness of the parallel algorithm.3.Deeply studied the basic principles of BP algorithm,listed the specific implementation steps of BP algorithm in detail,and analyzed the amount of calculation and parallelism of BP algorithm.Based on the CUDA architecture,a parallel BP imaging algorithm is designed,specific design ideas and implementation steps are given,and related optimization and analysis are performed.A simulation experiment was carried out to verify the accuracy and efficiency of the parallel algorithm.4.Research two sidelobe suppression algorithms: traditional spatial apodization and polynomial filtering and image fusion algorithms,and compare them.Based on the algorithm of polynomial filtering and image fusion,a parallel sidelobe suppression algorithm is designed,and the parallel design idea and specific implementation process are given.The kernel function is optimized by using shared memory to improve computing performance.Related simulation experiments are carried out to verify the correctness and performance of the parallel algorithm.