Imaging Algorithm Of Synthetic Aperture Radar Based On Graphics Processing Unit

With the extending of Synthetic Aperture Radar (SAR) application, it is required that SAR system can generate satisfactory image results under multiple modes, with a high resolution, and robust in rigid environment. All of these requirements bring new challenges to SAR signal processing.The complexity of the imaging algorithm should increase towards these challenges, but the processing time of the imaging algorithm has to decrease to produce real time SAR images. Therefore, we have to explore new computation platforms of SAR imaging algorithms and do research on efficient and high-performance SAR imaging algorithms which are suitable for these platforms.Recently, Graphics Processing Unit (GPU) is developing at top speed, which is a novel and promising computation platform for efficient SAR imaging algorithms. GPU is invented to solve the problems of highly complicated graphics processing computations in 3D era. Compared to CPU, general purpose computations based on GPU show higher performance with lower cost.The main purpose of this paper is to propose SAR imaging algorithms based on GPU. These new algorithms are suitable for GPU's architecture and programming model, take full advantage of GPU's computation resources and are more efficient compared to imaging algorithms based on CPU.At the beginning of this paper, the classic CPU-based SAR imaging algorithms are summarized, as theoretical basis of the research. And then, the developing status, advantages and limitations, and programming environment of GPU general purpose computation are introduced. After that, the implementation of SAR imaging algorithms based on GPU is discussed with details. Novel works done in this paper are summarized as follows.First, a SAR imaging algorithm based on GPU via OpenGL and Cg is proposed in this paper. It leverages the portability of OpenGL and Cg, so it can be used in different GPU platforms.Second, a SAR imaging algorithm based on GPU via CUDA is proposed in this paper. The implementation of this algorithm uses the latest GPU's developing environment, which takes full advantage of GPU's computation capability.Third, using texture mapping to implement range cell migration correction (RCMC) is proposed. Texture mapping is an important concept in computer graphics, which can use GPU hardware acceleration. The two GPU-based algorithms introduced in this paper both use texture mapping to implement RCMC.Last, a framework of GPU general purpose computation is introduced. In order to use OpenGL and Cg to implement GPU-based imaging algorithm, comprehensive knowledge of GPU's pipeline and programming interfaces is needed. A useful developing framework of GPU general purpose computation is designed for convenient development, by abstracting some programming interfaces of OpenGL and Cg, and proposed in this paper.The simulation results show, results of these two GPU-based algorithms introduced in this paper agree with theoretical values.The processing time of actual SAR data show, these two GPU-based algorithms are more efficient than traditional algorithms based on CPU. Especially, the imaging algorithm via CUDA is more than ten times as fast as CPU-based SAR imaging algorithms.The GPU-based imaging algorithms of synthetic aperture radar introduced in this paper show high efficiency and high performance. They provide a promising way to solve the problems of SAR signal processing in the future.