利用最小交叉熵的多时相影像变化检测  

Change detection of multi-temporal remote sensing images using minimum cross-entropy

作　　者：黄克凤 於雪琴 黄亮[2] 

机构地区：[1]昆明市测绘研究院,昆明650051 [2]昆明理工大学国土资源工程学院,昆明650093

出　　处：《测绘科学》2016年第1期99-102,共4页Science of Surveying and Mapping

基　　金：云南省教育厅科学研究基金项目(2013J062);云南省博士研究生学术新人奖支持项目

摘　　要：针对目前多时相遥感影像变化检测中缺少有效的确定变化阈值方法的问题,文章提出一种基于最小交叉熵的方法:采用中值滤波方法分别消除两个时相遥感影像中的噪声;将比值法和差值法融合构造两个时相遥感影像的差异影像;通过最小交叉熵确定差异影像的最佳变化阈值,并对差异影像进行分割得到变化区域;最后对变化检测精度进行评估。实验结果表明,提出的方法具有较高的检测精度,是一种有效的多时相遥感影像变化检测方法。Aiming at the insufficiency of efficient determination on the change threshold in multi-temporal remote sensing image change detection,the paper proposed an algorithm based on minimum cross-entropy：first,the noise in bi-temporal remote sensing images was eliminated respectively by median filtering method;and the ratio method merged with the difference method was used to obtain the difference image for the multi-temporal remote sensing image;then the best change threshold of the difference image was determined by minimum cross entropy method,and the difference image was segmented to obtain the change regions;finally,the accuracy of change detection was evaluated.Experimental result showed the feasibility of the method with high precision.

关 键 词：变化检测 遥感影像 最小交叉熵 影像分割 

分 类 号：TP75[自动化与计算机技术—检测技术与自动化装置]