双稀疏表示的遥感图像变化检测  

Remote Sensing Image Change Detection Based on Double Sparse Representation

作　　者：李骥[1] 肖雷鸣 王威[1] 

机构地区：[1]长沙理工大学综合交通运输大数据智能处理湖南省重点实验室,长沙410114

出　　处：《小型微型计算机系统》2018年第3期596-599,共4页Journal of Chinese Computer Systems

基　　金：国防"九七三"基金项目(613XXX0301)资助;湖南省教育厅科研课题项目(17C0043)资助

摘　　要：本文提出了一种双稀疏表示的变化检测方法以提高检测的精度,增强图像灰度鲁棒性.首先,对图像进行预处理,避免受到相干斑噪声的干扰.然后,将两幅时相图片进行双稀疏表示,利用不同的两个字典进行两次稀疏,以获得只包含有用信息的重构图像,以此作为后续处理步骤的输入图像.接下来利用Mean-Shift方法对图像进行分割提取特征,利用回归法构造差异图像.最后,对差异图像做阈值分割处理得到检测结果.实验结果表明与基于纹理特征的变化检测算法、UDWTKEAN算法和NSCTKFCM算法相比,本文所提出算法由于使用了双稀疏变换提高了检测的精度.In this paper, a change detection method based on double sparse performance is proposed to both improve the detection accu- racy and enhance the gray level robustness of image. First, the image is pretreated to avoid interference by speckle noise. Then, two temporal pictures are represented by double-sparse approach, and two sparse representations are made by two different dictionaries to obtain the reconstructed image containing only useful information as the input image of the subsequent processing step. Next, the Mean-Shift method is used to extract the features of the images, and the difference image is constructed by using the regression meth- od. Finally, the threshold image segmentation is performed on the difference image to obtain the detection result. The experimental re- suits show that compared with the texture detection algorithm based on Texture, UDWTKEAN algorithm and NSCTKFCM algorithm, the algorithm proposed in this paper improves the accuracy of detection by double sparse transformation.

关 键 词：遥感图像 变化检测 双稀疏表示 图像分割 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]