The Research Of SAR Imaging Algorithms Based On GPU

Synthetic Aperture Radar(SAR)has characteristics of all-time and all-weather of the earth observation without affecting by the conditions such as illumination and climate,and has been widely used in the civil field and military field.With the development of SAR technology,the resolution higher and higher,which makes the SAR echo data volume larger,and the imaging algorithm compute complexly.Use traditional CPU to deal the data is time-consuming,low efficiency,high cost of hardware.These bring a huge challenge to SAR real-time imaging processing.So the parallel optimization of SAR imaging processing based on high performance computing platform has become the hotspot of current research.In recent years,with the rapid development of GPU,its powerful floating point computing power provides a new computing platform for the large-scale parallel computing.More and more people use the CPU and GPU to mixed programming,making GPU-based parallel programming to be used in various fields.The study found that the application of GPU in the SAR data processing to accelerate the processing of SAR data can improve the current difficulties effectively.This thesis mainly studies SAR imaging algorithms based on GPU,and the main work and research results are as follows:1.This thesis researches the positive side view SAR imaging principle and SAR imaging algorithm in strip mode,which provides a theoretical basis for SAR parallel computing on GPU.At the same time,basing on the analysis of the development of SAR technology,the CUDA platform and its related technologies are studied,which can solve the complex computing and lay the technical foundation for parallel optimization of SAR imaging.2.Echo simulation is researched,in order to obtain specific echo data to verify the correctness of the SAR imaging algorithm on the GPU and CPU.First,simulation in the MATLAB to verify,and then achieve multi-point target simulation on the CPU and GPU to get the required echo data,finally analyze and compare the result.3.The parallel optimization technology of CS imaging algorithm based on GPU is researched.First,the CS algorithm is analyzed,and the algorithm of GPU parallelism is designed.Then the CUDA programming is used to realize the CS algorithm to obtain correct compression result,which is optimized by CUDA related technology.Compare with the processing time on the CPU platform,the speedup can reach up19.8 times.4.This thesis uses GPU to accelerate the CS algorithm in MATLAB.First,several common methods of using GPU acceleration in MATLAB areanalyzed and compared.Then combining with the basic flow of CS imaging algorithm,this thesis applies the build-in function of MATLAB to accelerate the algorithm and obtain better acceleration effect.