基于后向投影的SAR成像算法与GPU加速研究

发布时间：2018-04-02 04:34

  本文选题：合成孔径雷达　切入点：后向投影　出处：《南京航空航天大学》2014年硕士论文

【摘要】：合成孔径雷达(Synthetic Aperture Radar,SAR)是一种全天时、全天候的微波成像系统,高分辨率的特点使它在军用和民用领域有着不可替代的作用。随着合成孔径雷达成像技术的发展,各种高分辨率成像算法应运而生。然而高分辨率带来的巨大计算量成为某些成像算法实际应用的瓶颈,其中最为典型的就是后向投影(Back Projection,BP)算法。BP成像算法是一种时域成像算法,与传统SAR成像算法相比,该成像算法原理简单,并且在原理上不存在任何理论近似,能够实现高分辨率SAR成像。因此,更具有实际的研究价值。基于以上背景,本文的主要工作如下:(1)分析了BP成像算法以及快速BP(Fast Back Projection,FBP)成像算法的成像模型、实现原理和计算量。针对BP成像算法计算量巨大的特点,本文在BP算法的基础上实现了一种FBP成像算法。实验结果证明该FBP成像算法与BP成像算法成像质量相当,并且FBP算法在一定程度上从算法层面降低了计算量。(2)研究了BP成像算法在非理想航迹下的运动误差以及运动补偿技术。分析了BP成像算法误差的来源,建立运动误差模型,分别提出了基于对比度最优准则的自聚焦方法以及基于划分子孔径和对比度准则的BP成像自聚焦方法,实测数据成像结果验证了这两种运动补偿算法可分别有效应用于短孔径和长孔径成像处理。(3)介绍了基于计算统一设备架构(Compute Unified Device Architecture,简称CUDA)环境下的图形处理器(Graphic Processing Unit,GPU)编程技术,分析了BP成像算法的内在并行性,提出了一种适合GPU加速实现的BP成像算法加速方案;针对SAR处理数据量较大以及GPU显存受限的问题,在此方案的基础上进一步提出基于流技术的GPU优化方案。实测数据处理结果为优化后比优化前平均成像速度提升约78.8%,表明该方案的有效性和可行性。
[Abstract]:Synthetic aperture radar (Synthetic Aperture, Radar, SAR) is a kind of all day long, microwave imaging system of all-weather, high resolution. It plays an irreplaceable role in military and civilian fields. With the development of synthetic aperture radar imaging technology came into being, high resolution imaging algorithm. However, large amount of calculation and high resolution the practical application has become a bottleneck of certain imaging algorithms, the most typical is the back projection (Back Projection BP) algorithm for.BP imaging algorithm is a time domain imaging algorithm, compared with the traditional SAR imaging algorithm, imaging principle of this algorithm is simple, and there is no theory in principle, can achieve high resolution SAR imaging. Therefore, the research has more practical value. Based on the above background, the main work of this paper are as follows: (1) the analysis of BP imaging algorithm and fast BP (Fast Back Projection, F BP) model imaging algorithm of imaging principle and calculation. According to BP imaging algorithm computation is huge, this paper implements a FBP imaging algorithm based on BP algorithm. The experimental results show that the FBP imaging algorithm and BP imaging algorithm for image quality, and the FBP algorithm to a certain extent from algorithm the level of the computation is decreased. (2) the motion error and motion compensation of BP imaging algorithm under the nonideal track. The sources of error of BP imaging algorithm, establish the kinematic error model are proposed based on the quasi contrast optimization autofocus method and BP imaging sub aperture and contrast criterion based on the self focusing method, real data imaging results verify that the two kinds of motion compensation algorithm can be effectively applied to respectively short aperture and long aperture imaging processing. (3) is introduced based on Compute Unified Device Architecture (Compu Te Unified Device Architecture, referred to as CUDA) graphics processor environment (Graphic Processing Unit GPU) programming technology, analyzes the inherent parallelism of BP imaging algorithm, proposes a BP imaging algorithm for GPU GPU acceleration scheme for SAR data processing; and a large amount of GPU memory constrained problem based the scheme is further proposed GPU optimization scheme based on streaming technology. The processing results of real data for optimization than before optimization average imaging speed increases about 78.8%, shows the validity and feasibility of the scheme.

【学位授予单位】：南京航空航天大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TN957.52
，

本文编号：1698827