基于GPU的阵列雷达信号处理技术研究

发布时间：2018-04-30 05:35

本文选题：阵列雷达 + 信号处理　；参考：《电子科技大学》2017年硕士论文

【摘要】：随着雷达信号处理算法的研究不断深入,信号处理系统的复杂性与实效性的矛盾更加突出,提高雷达信号处理速度成为必要。近年来,GPU(Graphics Processing Unit,图形处理器)技术的推广和CUDA(Compute Unified Device Architecture,计算统一设备架构)并行编程的推出,使研究人员找到了提高雷达信号处理效率的新方向,研究发现雷达信号处理的大量算法都具有并行特点,可利用GPU并行实现来加速处理过程。本文以阵列雷达信号处理系统的GPU实现为背景,主要工作如下:1、介绍了GPU通用计算模型,分析了GPU的基本硬件架构和CUDA编程模型,并总结了几个CUDA编程的优化方向。2、研究了阵列雷达信号处理系统的总体框架,确定了系统的工作流程,然后结合项目需求设计了系统各模块的算法和具体实现步骤,并在CPU(Central Processing Unit,中央处理器)上运用C语言编程实现,最后利用MATLAB仿真数据验证了各模块的功能,完成了整个软件化系统的CPU实现。3、研究阵列雷达信号处理系统的GPU实现。首先分析了系统各模块实现算法的并行结构,完成了各模块的并行算法设计;然后利用CUDA编程在GPU上实现,并利用仿真数据验证了各模块实现的正确性;再通过对比CPU的处理速度,分析了各模块的GPU加速效果与性能限制因素,提出了相关优化策略;最后运用实测数据验证了该系统能够正确处理雷达回波数据,并且在本实验平台下该系统在GPU上的加速比能达到104倍左右,具有良好的加速效果。4、完成了整个阵列雷达信号处理系统的软件调试,并给出了系统联调结果,处理结果表明该系统能够稳定地观测目标运动轨迹,满足系统需求。
[Abstract]:With the development of radar signal processing algorithm, the contradiction between complexity and effectiveness of signal processing system is becoming more and more serious, so it is necessary to improve the speed of radar signal processing. In recent years, the popularization of GPU Graphics Processing Unit (GPU) technology and the introduction of parallel programming in CUDA(Compute Unified Device Architecture, computing unified equipment architecture have led researchers to find a new direction to improve the efficiency of radar signal processing. It is found that a large number of radar signal processing algorithms have parallel characteristics and can be implemented in parallel with GPU to speed up the processing process. In this paper, the GPU implementation of array radar signal processing system is taken as the background. The main work is as follows: 1. The general computing model of GPU is introduced, and the basic hardware architecture and CUDA programming model of GPU are analyzed. The optimization direction of several CUDA programming is summarized. The general framework of array radar signal processing system is studied, and the working flow of the system is determined. Then, the algorithm and implementation steps of each module of the system are designed according to the requirements of the project. Finally, the function of each module is verified by using MATLAB simulation data, and the CPU realization of the whole software system is completed. The GPU implementation of array radar signal processing system is studied. Firstly, the parallel structure of each module is analyzed, and the parallel algorithm design of each module is completed. Then, the parallel algorithm is implemented on GPU by using CUDA, and the correctness of each module is verified by simulation data. By comparing the processing speed of CPU, the acceleration effect and performance limitation factors of each module are analyzed, and the relative optimization strategy is put forward. Finally, it is verified that the system can process radar echo data correctly by using the measured data. The speedup ratio of the system on GPU is about 104 times under this experimental platform, and it has good acceleration effect .4. the software debugging of the whole array radar signal processing system is completed, and the result of the system combination is given. The processing results show that the system can stably observe the moving track of the target and meet the needs of the system.
【学位授予单位】：电子科技大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TN957.51

【参考文献】