【作者】 朱朝杰；

【导师】 王仁礼；

【作者基本信息】 解放军信息工程大学 ， 环境工程， 2007， 硕士

【摘要】 图像变化检测技术旨在检测同一地区不同时相的变化情况。这一技术在民用与军事等方面都有着非常广泛的应用,其多年的理论研究和具体实践也己产生了大量的算法。当前基于遥感影像的变化检测算法很多是以像元为基础进行研究,而从纹理和形状等模式来综合研究相对较少。本文从目标图像特征分析着手进行遥感影像的变化检测技术研究与探讨,工作之一是利用小波理论进行图像纹理特征分析,然后进行变化检测;工作之二是利用目标的不变特征进行形状分析与提取,进而应用于变化分析。本文的主要工作和创新点如下:1)阐述了变化检测研究背景和现状,指明了当前存在的主要问题,明确了本文的研究范围和基本思路。2)概述出变化检测的一般流程,对用于变化检测的数据源以及图像预处理,如几何校正、辐射校正等关键技术进行了分析,并根据应用对象,即面状、线状及独立目标,对目前算法作了简要归类,详细阐述了各类算法中较经典的一些算法。3)研究并实现了基于纹理特征的小波变化检测方法。该方法引入了小波理论,对其低频和高频系数采取不同的纹理特征提取方法,与传统方法相比具有更好的检测效果,比直接提取纹理特征又具有更快的速度,提高了检测效果与效率。4)以传统迭代算法所获得的阈值为基础,增加一适当的常量,使其更适合于分割目标在影像中占有比例较小而且灰度较亮的情况,通过与其他方法比较得出其具有算法简单、分割效果更好的结论。5)针对目标形状特征设计并实现了基于多不变特征的目标提取及变化检测方法。该方法主要针对独立目标(如人造小面元目标),实验表明其检测精度高且可实现各自影像变化目标的定位分析。更多还原

【Abstract】 The technology of change detection based on remotely sensed images means that given the remotely sensed multi-temporal images, panchromatic or multi-spectral, one’s aim is to detect the changes of ground feature or target by means of image processing and analyzing. This technology plays an important role in the field of civil and military affairs, and a large volume of algorithms have been put forward. Most of these algorithms were studied based on pixels, while only a few are synthetically studied from other patterns of image such as texture and shape. The aim of this paper is to realize the change detection of ground feature or target in remotely sensed images by image feature analysis. On the one hand, using image texture feature analysis by wavelet transformation is to detect change of images; on the other hand, using feature invariant of target image is to analyze the shape of targets, then extract targets and detect those changes. Main works of this paper are as follows:1) It analyzes the background and current status of change detection, points out the main problems of the existing algorithms, brings forward the range and basic ideas of this paper.2) Aiming to common process of change detection based on remotely sensed multi-temporal images, it introduces the data source and the work of image preprocessing in favor of change detection, such as geometric correction and radiometric correction. And according to the applied objects, it summarizes the existing algorithms, detailed expounded some classical methods.3) It studies an algorithm of change detection based on texture feature by wavelet. This method applies wavelet theory to extract different features from wavelet approximate constituent and detail constituents. Compared to conventional methods, this method has better effect, and has faster speed than directly extracting texture feature. So it improves change detection effect and efficiency.4) The segmentation algorithm of iterative threshold is developed, which is rather suitable to segment targets with small proportion of the image. And compared to other existing methods, this method is simple and has better effect of segmentation.5) This paper also puts forward a method of target extraction and its change detection based on multi-invariant features. It mainly aims at the isolated target, realizing the quantitative and positional analysis with fine accuracy for mutative targets.更多还原