相似性判别原理在影像变化检测中的应用

Application of Similarity Discriminating Theory in Change Detection of Images

针对传统基于Bayes影像变化监测方法中分类的硬划分及对辐射畸变敏感的问题,首先,采用集合相似性测度动态更新变化和未变化两类像元模糊子集,解决变化检测分类过程中硬划分的问题;然后,采用像元相似性测度来判断变化类像元集合和未变化类像元集合中的每个像元是否真正发生了变化,剔除伪变化像元,克服对影像线性辐射畸变敏感的问题,并检测出漏检像元,最终实现影像变化区域的提取。实验结果表明,基于相似性测度的影像变化检测方法可实现变化区域自动化提取,具有较好的应用价值。

Since there is deficiency in classification and sensitivity to radiometric distortion of traditional change detection approach of images based on Bayes decision rules,at first, changed pixels subset and non-changed pixels subset are updated dynamically depending on set similarity, and the problem of classification in change detection of images is solved. Then it analyzes each pixel of changed pixels set and unchanged pixels set to find the non-changes depending on pixel similarity, and it is effective to solve the problem of sensitivity to radiometric distortion of image ,and to detect residual pixels; At last, the changed regions from images are extracted. Through experiments, it proves that the change detection approach of images based on similarity discriminating analysis can extract the changed region quickly and accurately, which has preferable application value.