基于OpenCL的并行SAR图像变化检测

发布时间：2018-06-18 23:03

本文选题：变化检测 + SAR　；参考：《西安电子科技大学》2014年硕士论文

【摘要】：SAR(Synthetic Aperture Radar)图像变化检测是通过检测同一地区不同时段获取到的SAR影像,进而发现其地表发生的变化信息。图像的变化检测已经广泛地应用在诸多国民经济和国防建设中,比如在环境监测、自然灾害评估、军事侦察以及战场信息动态感知等领域。由于这些应用领域都需要做出快速的反应和及时的决断以及海量SAR图像处理带来的庞大的计算挑战,而传统的基于单CPU的处理方法难以满足和应对这些挑战。这时异构计算正作为一个新型的计算机体系迅猛发展并逐渐成为了主流,异构计算的迅猛发展为并行算法的设计和实现提供了可能和前提,所以需要利用已有的并行计算设备、软件资源和并行的分析技术来应对这些挑战,解决这些问题。因此,本文就并行SAR图像变化检测进行了深入而又系统的研究,为了验证所提出算法的性能,在众多异构加速设备上对所提出的算法进行了验证,并将此算法利用Qt封装成软件系统,本文的主要工作如下:(1)本文将快速全局K-均值(Fast Global K-Means,FGKM)算法用于SAR图像变(1)本文将快速全局K-均值(Fast Global K-Means,FGKM)算法用于SAR图像变化检测之中,其实验结果显示此算法可以获得良好的效果,但FGKM算法比较耗时,因此本文提出了一种基于并行快速全局K-均值(Parallel-Fast Global K-Means,P-FGKM)的SAR图像变化检测算法,该算法是基于Open CL(Open Computing Language)框架实现的,由于Open CL框架的可移植性,因此,我们在多核CPU、单GPU(Graphics Processing Unit)、多GPU等异构加速设备上对所提出的算法进行了验证,并对其结果进行理论分析。(2)为了研究基于Open MP(Open Multi-Processing)框架和基于Open CL框架的并行算法的差异,本文还研究了基于Open MP框架的并行SAR图像变化检测算法,在多核CPU、MIC(Many Integrated Core)卡上对算法进行了测试,并分析其结果。此外,本文还进行了并行SAR图像变化检测系统软件的设计与实现,将变化检测算法利用Qt进行了封装。具体工作是进行软件UI界面的设计与布局,进行信号/槽的连接等问题,形成了一套软件界面,使得用户便捷地利用此软件进行SAR图像的并行变化检测。
[Abstract]:Sar synthetic Aperture Radar image change detection is based on the detection of SAR images obtained in different periods of time in the same area, and then the change information on the surface of the SAR image is found. Image change detection has been widely used in many fields of national economy and national defense, such as environmental monitoring, natural disaster assessment, military reconnaissance and battlefield information dynamic perception. These applications require rapid response and timely decision, as well as huge computational challenges brought by massive SAR image processing, but the traditional single-CPU processing methods are difficult to meet these challenges. At this time, heterogeneous computing is developing rapidly as a new computer system and has gradually become the mainstream. The rapid development of heterogeneous computing provides the possibility and premise for the design and implementation of parallel algorithms, so it is necessary to use the existing parallel computing equipment. Software resources and parallel analytical techniques are used to address these challenges and solve these problems. In order to verify the performance of the proposed algorithm, the proposed algorithm is verified on many heterogeneous acceleration devices. This algorithm is encapsulated into a software system using QT. The main work of this paper is as follows: 1) in this paper, Fast Global K-Means-FGKM (Fast Global K-Means-FGKM) algorithm is applied to SAR image variation. In this paper, Fast Global K-Means-FGKM algorithm is applied to SAR image change detection. The experimental results show that the algorithm can achieve good results, but the FGKM algorithm is time-consuming. Therefore, this paper proposes a change detection algorithm for SAR images based on parallel fast global K-Means-Fast Parallel-Fast K-Means-P-FGKM. The algorithm is based on Open CLU Open Computing language framework. Because of the portability of Open CL framework, we validate the proposed algorithm on heterogeneous accelerating devices such as multi-core CPU, single GPU Graphics processing Unit and multi-GPU. In order to study the differences between parallel algorithms based on Open MPU Open Multi-Processing-based framework and Open CL framework, this paper also studies the parallel SAR image change detection algorithm based on Open MP framework. The algorithm is tested on the multi-core CPUU MICMany Integrated Core Card and the results are analyzed. In addition, the software of parallel SAR image change detection system is designed and implemented, and the change detection algorithm is encapsulated by QT. The main work of this paper is to design and layout the UI interface of the software and to connect the signal and slot. A set of software interface is formed, which makes the user use the software to detect the parallel changes of SAR images conveniently.
【学位授予单位】：西安电子科技大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TN957.52

【相似文献】