基于多时相PCA光谱增强和多源光谱分类器的SPOT影像土地利用变化检测  

Detecting Land Use Change Using PCA-Enhancement and Multi-Source Classifier from SPOT Images

作　　者：邓劲松[1] 李君[2] 王珂[1] 

机构地区：[1]浙江大学农业遥感与信息技术应用研究所,浙江杭州310029 [2]浙江省环境保护科学设计研究院,浙江杭州310007

出　　处：《光谱学与光谱分析》2009年第6期1627-1631,共5页Spectroscopy and Spectral Analysis

基　　金：国家“863”计划项目(2006AA10Z204);国家自然科学基金项目(3080070);国家博士后基金项目(20070421194)资助

摘　　要：在全球快速城市化的大背景下,土地利用变化检测始终是全球变化研究的重点和热点。文章研究利用2003和2006年高分辨率SPOT-5遥感影像,在进行高精度的正射纠正后,运用多时相PCA光谱增强和多源光谱分类器相结合的方法进行城市土地利用变化检测。结果表明,多时相PCA光谱增强后得到前3个主成分集中了绝大部分光谱信息,其中PC1和PC2增强了土地利用未发生变化的光谱信息,而变化信息主要集中在PC3。而多源光谱分类器准确地提取出各种变化和未变化信息。精度评价结果表明,本文提出的变化检测方法的总体精度达到92.58%,Kappa系数为0.92,用户精度和生产精度也都取得满意的结果,并且精度都明显高于常规的方法(分类后比较法)。Concomitant with the rapid global urbanization process, land use change detection has been the focus and ＂hot spot＂ of global change research all the time. In the present study, the rigorous orthorectification was first applied to the SPOT-5 data to guarantee precise geometric correction and image registration. Afterwards, a methodology integrating PCA-enhancement and multi-source classifier was adopted to detect the land use changes in urban area. The results show that the first three PCs derived from multi-temporal-PCA contain most of the spectral information among which unchanged land use is highlighted in PC1 and PC2, and changed land use is mainly enhanced in PC3. The following multi-source classifier integrating unsupervised classifier （ISODATA） and supervised classifier （Maximum Likelihood） accurately extracts all the information. The findings from accuracy assessment demonstrate that the overall accuracy for the proposed method reaches 92.58%, KAPPA coefficient is 0. 92, and proving figures are also produced for the user＇s and producer＇s accuracies. It was further found that the proposed method yielded better accuracy than that of traditional post-classification comparison approach.

关 键 词：SPOT-5 主成份分析 土地利用 变化检测 

分 类 号：S28[农业科学—农业水土工程]