Research And Implementation Of GPU-based Shooting-and-bouncing Ray Method Electromagnetic Scattering Simulation And Visualization Platform

In the field of microwave remote sensing,stealth and anti-stealth on the background of sea surface,the main research object is the electromagnetic scattering of sea surface and the electromagnetic scattering of the composite model of sea surface and targets.In the study of the electromagnetic scattering of sea surface,a sea surface is regarded as a random rough surface.Furthermore,random rough surfaces often are regarded as electrically large targets.In practice,high-frequency methods are usually employed to implement fast electromagnetic scattering computation of electrically large targets.These high-frequency methods,including Physical Optics(PO)and Shooting and Bouncing Ray(SBR),are simple to implement and highly parallelistic.Due to the fact that CPU usually contains only a few physical processors,it is difficult to take advantage of the parallestic nature of the high-frequency method.GPU,however,usually contains hundreds and thousands of physical processing units.There is potential for making full use of all GPU processing units to utilize the the parallelism of high-frequency method.That is premise of this thesis.This thesis revovles around the following objectives:First of all,the fundumentals of Geometrical Optics(GO),Physical Optics(PO),Shooting and Bouncing Ray(SBR)and PO-SBR method is investigated in detail.Because ray tracing is an essential part of Shooting and Bouncing Ray,the ray tracing process is also investigated in detail.Secondly,the fundumentals of GPU parallel programming and concepts of CUDA parallel programming platform are introduced.The discussion includes CUDA programming model and many architectural differences between GPU and CPU.The discussion further extends to GPU instruction pipeline,GPU memory model and CPUGPU communication via PCIe interface.In GPU parallel programming,strategies used to allocate of computing tasks,make full use of both CPU and GPU and reduce the CPUGPU communication latency,are the core of GPU parallel algorithm.Subsequently,the implementation of parallel PO-SBR method based on CUDA parallel programming platform is discussed.In the process,several ray tracing algorithms are studied and their strength and weakness are analyzed.In addition,visualization of electromagnetic simulation results is a common need for the study of electromagnetic scattering.This thesis independently developed a crossplatform graphics engine based on OpenGL API that can be used for both electromagnetic scattering visualization and general graphics programming.This thesis used this graphics engine to implement the visualization of induced current on surface of targets.At this point,the algorithm implementation of GPU based PO-SBR method and the platform implementation form a tool chain that could be used to study high-frequency method more effectively.Finally,several test cases and test case studies are provided for verification of the accuracy and efficiency of algorithm implementation and platform implementation.