SAR Fast Imaging And Target Detection Methods And GPU Implementation

Synthetic aperture radar(SAR)is not affected by time,weather,geographical and other factors.It has many advantages such as all-weather work and long operational distance.SAR is widely used in target detection,identification and location,battlefield reconnaissance and surveillance,prediction of natural disasters,resources exploration and other military and civilian fields.SAR can get high resolution image,but the imaging algorithm is complex,and it need to deal with a large number of echo data which makes the amount of computation very large,and puts forward high demands on the hardware platform.According to the application requirements of spaceborne SAR imaging,this paper carried out the research of SAR fast imaging,target detection,and imaging evaluation,its main contents include:(1)The design factors of spacebome SAR signal processing system are analyzed,and an onboard SAR signal processing system based on embedded GPU is given,the efficiency of GPU parallel processing is analyzed,and the GPU programming model and storage model of CUDA platform is discussed.(2)The SAR echo model is analyzed,the storage method of SAR echo signal is designed;The Chirp Scaling(CS)imaging algorithm has been optimized for parallel processing structure.The CS parallel imaging algorithm is designed based on the GPU signal processing platform;The experimental results show that compared to the CPU platform,the system achieved dozens or even hundreds of speed ratio.(3)The two parameter CFAR algorithm has been optimized for parallel processing structure and the two parameter CFAR algorithm of parallel program is designed.The SAR images of real ship is used in verification.The results show that the method is accurate,and the acceleration ratio is more than 400 compared with the CPU platform.(4)The quality objective evaluation index of SAR image is discussed,the GUI test software of SAR image quality assessment is designed by C#,and the quality of SAR images gotten from the spaceborne SAR fast imaging platform and the traditional CPU imaging platform are compared and analyzed.