A GPU Based Real-time SAR Simulation For Complex Scenes

As a microwave remote sensing tool, Synthetic Aperture Radar has all-sunlight and all-weather features. It has great potential applications in the fields of military and civilian. However, compared to the optical images, SAR images have three typical characteristics, which are reflection properties, geometrical aberrations and speckle noises. It brings difficulties to analyze images, to get geometrical orientation and to identify the targets. The simulation of a special target's image in a given condition expands a new way to collect target's samples, to predict target's features and to implement target's recognition. The conventional SAR image simulation is based on the platform of CPU. It has many defects, such as time-consuming, not intuitionistic, lack of interaction and so on, especially in the simulation of complex scenes.Recently, with the rapidly increasing processing power and the growing levels of the programmability of graphics processing units (GPU), the applications of GPU to computer graphics have become a hot spot of research. Especially in the industry of movies and games, it brought about extensive applications and enormous economic effects. And in turn, it promotes the development of GPU. Now, more and more researchers begin to introduce the GPU to do kinds of intensively computational work, in order to accelerate the computation processing by using the streamlined and parallelizable method of GPU.This dissertation presents a novel simulation to generate SAR images for complex scenes. It introduces the techniques of computer graphics to SAR simulation and makes full use of 3D hardware acceleration and programmable GPU, which is offered by graphics hardware. It implements real-time and effective SAR simulation. The main jobs and innovations of the dissertation are as follows:①Constructing the large-scale terrain scenes, which is the basis of SAR simulation. Real-time interactive rendering of terrain scene is a great challenge to graphics hardware due to its massive data. So in order to improve the rendering frame rate, we have to push as much triangles through the pipeline of GPU, which can relieve the burden of CPU and make good performance of GPU.②Using the Phong lighting model to simulate the backscatter characteristics of plane facets. The reflective coefficients of triangles in the 3D terrain model are corresponding to backscatter coefficients of facets. According to the roughness of the material surface, the reflection is either more Lambertian or more specular.③Calculating the real-time shadow in the SAR simulation. In the conventional SAR simulation of geometrical distortion, it usually calculating the shadow by point-to point comparison, which is time-consuming. The dissertation adopts the shadow mapping method in computer graphics to compute the shadow quickly. It is implemented by giving the time-consuming job to graphics hardware.The dissertation uses OpenGL (3D API function) and Cg (advanced shading language) to programme on the hardware platform of NVIDIA Quadra FX3700. OpenGL is a 3D graphics software interface, which is developed by SCI Company. It is easily transplantable. Cg is a special programme language for GPU. It is developed by NVIDIA and Microsoft, and works with OpenGL together. The experimental results show that, the simulation is real-time and effective. It can make full use of programmable pipeline of GPU and relieve the burden of CPU effectively.