High-resolution Radar Image Simulation Of Complex Targets Based On Fast GPU Computation

As an active microwave sensor,synthetic Aperture Radar(SAR)is widely used in remote sensing earth observation.The design of the SAR system,imaging algorithm,and image interpretation requires a large number of radar images with different attitudes in different imaging scenarios.However,due to the limitation of objective conditions,it is time-consuming and labor-intensive to acquire an enormous Radar image datasets,which is not easy to update and maintain.Hence,it is imperative to generate simulations of radar images under specific imaging conditions and parameter settings.Signal-based radar image simulation can better meet the application requirements of SAR image interpretation and target identification.How to quickly and accurately calculate the backscattering field of the target has been the research hotspot.However,the implementation of physical optics(PO)and geometric optics(GO)methods need to perform shadowing tracking,which requires heavy computational time.Besides,the computational accuracy needs to be improved.Therefore,this thesis develops a SAR image simulation method for complex targets based on GPU parallel processing.This method combines the Shooting and Bouncing Ray(SBR)algorithm and Graphics Processing Unit(GPU).The SAR echo signals are generated by calculating multiple scattering fields and antenna pattern tracking.The main works of the thesis are as follows:1.InSAR data and signals,we describe the simple pulse,pulse sequence,linear FM,and FM continuous wave signal models.The conversion relationship of time and space coordinate systems between targets and satellites is explored,which helps solve the problems of platform motion and radar image focusing during SAR image simulation.Besides,we introduce three commonly-used SAR imaging algorithms,which will be applied to the later implementation of the fast algorithm.2.Given the high computational complexity of the SBR algorithm in dealing with electrically large-sized targets,we design a GPU-accelerated algorithm.The efficiency of the proposed algorithm in this paper is validated by comparing the computational complexity of the Central Processing Unit(CPU)and the GPU.Then,the efficiency of the proposed algorithm is verified by using regular targets,such as perfect conductors,dielectric spheres,and dihedral corner reflectors.3.Combined with GPU-based fast computation of the backscattering field and the rangeDoppler algorithm,we simulate the high-resolution radar images of the complex targets,including the aircraft,tanks,and complex scenes on the sea.Finally,the richer target feature information is obtained by the radar image’s polarization feature decomposition.The simulation results show the accuracy and efficiency of the simulation developed in this thesis.