基于多目标模糊聚类的SAR图像变化检测

发布时间：2018-04-13 18:30

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

【摘要】：最近一些年来,合成孔径雷达技术(Synthetic Aperture Radar,SAR)获得高速发展,星载合成孔径雷达系统已经持续观测地球表面几十年,并且已经获得了大量的多时相地面观测数据。很多研究遥感领域的学者都想要更好地利用这些由合成孔径雷达系统获得的信息来深入探索合成孔径雷达技术,其中包括:对象分类、干涉测量、目标提取、边缘检测以及图象变化检测等等。在所有这些研究中,关于SAR图象变化检测技术的研究是最受欢迎的技术研究之一。图象的变化检测是这样一种技术,它依据同样地域不一样时间的两幅图象之间的比较和分析,最终目的是找到图象之间发生变化的区域。它现在已经被人们应用到了各种各样的领域中,如:自然及生态环境的实时监测、自然灾害的受灾面积估计和农田内各种作物的分布状况调查等。本文的重心是基于对SAR图象变化检测算法的研究,针对现有SAR图象变化检测算法中普遍存在的问题提出新的解决方案,重点包含以下工作:1.针对SAR图象变化检测问题中所亟待解决的图像细节保持与噪声去除间的矛盾,对众多已有SAR图象变化检测算法进行分析,首次从多目标优化的角度来理解变化检测问题,分别将图象细节保持能力与噪声去除能力作为两个待优化的目标,然后同时进行优化,从而将SAR图象变化检测问题转化为一个多目标优化问题。2.在将SAR图象变化检测问题转化成一个多目标的优化问题进行求解的基础上,将模糊聚类的隶属度更新与MOEA/D相结合,利用拉格朗日乘子法计算出使得每个分解得到的子问题的代价函数达到最小值得必要条件,求出隶属度矩阵的更新公式。3.将以上提出的改进点应用到几组真实的SAR图象数据集上,并和其它几种已经存在的检测算法进行详细的对比分析和说明,从而验证本文几个改进点的实用性和有效性。
[Abstract]:In recent years, synthetic Aperture radar (SAR) technology has gained rapid development. Spaceborne SAR system has been continuously observing the Earth's surface for several decades, and has obtained a large number of multi-phase ground observation data.Many researchers in the field of remote sensing want to make better use of the information obtained by the synthetic Aperture Radar system to explore SAR technology in depth, including object classification, interferometry, target extraction.Edge detection and image change detection and so on.Among all these studies, the research on SAR image change detection is one of the most popular techniques.Image change detection is a kind of technique, which is based on the comparison and analysis of two images with different time in the same region, and the ultimate goal is to find the region of change between the images.It has been applied to various fields, such as the real-time monitoring of natural and ecological environment, the estimation of natural disaster area and the investigation of crop distribution in farmland, and so on.The center of gravity of this paper is based on the research of the SAR image change detection algorithm, and proposes a new solution to the common problems in the existing SAR image change detection algorithm. The emphasis includes the following work: 1.In view of the contradiction between image detail preservation and noise removal which is urgently needed to be solved in SAR image change detection, many existing SAR image change detection algorithms are analyzed, and the change detection problem is first understood from the point of view of multi-objective optimization.The image minutiae retention ability and noise removal ability are regarded as the two targets to be optimized, and then optimized simultaneously, thus the change detection problem of SAR image is transformed into a multi-objective optimization problem. 2.On the basis of transforming the problem of SAR image change detection into a multi-objective optimization problem, the membership degree updating of fuzzy clustering is combined with MOEA/D.By using the Lagrange multiplier method, the necessary and necessary conditions for minimizing the cost function of each decomposed subproblem are obtained, and the updating formula of membership matrix .3is obtained.The proposed improved points are applied to several real SAR image data sets and compared with other existing detection algorithms in detail to verify the practicability and effectiveness of several improved points in this paper.
【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】