综合环境卫星与MODIS数据的面向对象土地覆盖分类方法  

Land Cover Classification of Compositing HJ-1 and MODIS Data Based on Object-based Method!

作　　者：贾明明[1,2] 王宗明[1] 张柏[1] 丁智[1,2] 

机构地区：[1]中国科学院东北地理与农业生态研究所湿地生态与环境重点实验室,吉林长春130102 [2]中国科学院大学,北京100049

出　　处：《武汉大学学报（信息科学版）》2014年第3期305-310,共6页Geomatics and Information Science of Wuhan University

基　　金：中国科学院重点部署课题资助项目(KZZD-EW-08-02);中国科学院国家外国专家局创新团队国际合作伙伴计划资助项目(2012ZX07207-004)~~

摘　　要：使用面向对象方法对单时相的环境卫星数据进行土地覆盖分类时,几何特征和光谱特征相似的地物无法区分,而MODIS时序数据的空间分辨率较低,不适用于中小尺度的土地覆盖分类。应用面向对象方法,充分利用环境卫星数据的空间、光谱特征和MODIS数据的物候特征建立规则,进行分类,可以有效地解决上述困难。首先对环境卫星数据进行多尺度分割,生成待分类对象;再根据对象的特征,依据由简到难的原则进行分层分类。以双台子河口为例进行土地覆盖分类,总体精度93%,Kappa系数0.92。结果表明,综合环境卫星与MODIS数据的面向对象土地覆盖分类方法应用潜力巨大。Identification of land cover types, which have similar texture features and spectral features,is very difficult based on object-oriented classification method and single date images. Due to the rough spatial resolution, MODIS times series data are not suitable to classification of land cover at small scale. In this study, a classification method for composing HJ-1 and MODIS time series data based on an object-oriented method is presented. In this method, classification rules were constructed using textural, spectral features from HJ-1 data and phonological features from MODIS time series da- ta. During the classification process, objects are built by multi-scale segmentation, and then land cov- ers are extracted one by one. The overall accuracy of these classification results are 93% with a Kappa confidence of 0.92, which indicates the classification method presented in this study is effective

关 键 词：面向对象方法 环境卫星 MODIS时序数据 双台子河口 土地覆盖 

分 类 号：P237.9[天文地球—摄影测量与遥感]