基于PCA的变化向量分析法遥感影像变化检测  

Remote sensing image change detection based on change vector analysis of PCA component

作　　者：黄维[1,2] 黄进良[1] 王立辉[1] 胡砚霞[1,2] 韩鹏鹏 

机构地区：[1]中国科学院测量与地球物理研究所,武汉430077 [2]中国科学院大学,北京100049

出　　处：《国土资源遥感》2016年第1期22-27,共6页Remote Sensing for Land & Resources

基　　金：中科院战略性先导科技专项专题"应对气候变化的碳收支认证及相关问题"(编号:XDA05050107)资助

摘　　要：为实现对土地覆盖变化的遥感监测,研究了一种基于不同年份单时相遥感数据提取差异影像、自动确定变化阈值提取变化区域的新方法。以南通市Landsat8 OLI影像为例,对2期影像分别进行主成分分析(principal component analysis,PCA);取前3个主分量进行变化向量分析(change vector analysis,CVA),构造变化检测差异影像,并与传统PCA法和CVA法构造的差异影像进行对比;对3景差异影像分别用传统全局阈值法和局部最小错分概率法自动确定阈值,分别提取变化区域,得到6景变化区域图。利用目视解译样点进行精度评价的结果表明,改进后的基于PCA的CVA法提取的变化区域总体精度可达92.78%,Kappa系数可达0.842 6,证明使用该方法可有效地进行不同年份单时相遥感数据的变化检测。In order to monitor the change of land cover with remote sensing technology,the authors studied a method which is based on single- temporal remote sensing image in different years for extracting differences between the images and determining the change threshold automatically to extract the change area. The research took Landsat8 OLI images of Nantong City as an example. Principal component analysis（ PCA） was carried out respectively on two images. After the PCA transformation,the first three components were operated based on change vector analysis（ CVA） to get the difference image for change detection,which was compared with the extraction results based on the traditional PCA method and CVA method. Overall minimum error probability threshold determination method and local minimum error probability method were utilized to automatically determine the threshold of the three difference images and to get six change area images. The accuracy was evaluated by visual interpretation,and the results show that the overall accuracy of the new method can reach 92. 78%,with kappa coefficient up to 0. 842 6. This method is proved to be feasible and effective for extracting change area by single-temporal remote sensing image in different years.

关 键 词：变化向量分析(CVA) 主成分分析(PCA) 阈值确定 变化检测 

分 类 号：TP751.1[自动化与计算机技术—检测技术与自动化装置]