基于影像多种特征的决策树分类方法  

Method for Decision-tree Classifier Using Multi-feature of Images

作　　者：闫利[1] 孙颖超[1] 

机构地区：[1]武汉大学测绘学院,湖北武汉430079

出　　处：《地理空间信息》2009年第6期15-17,共3页Geospatial Information

基　　金：国家863计划资助项目(2007AA12Z154)

摘　　要：阐述了决策树分类CART算法原理,将纹理信息、NDVI指数引入决策树方法对影像进行分类,并将分类结果与最大似然分类结果进行比较,研究表明决策树分类方法相对传统分类方法总体精度提高了8.9148%,Kappa系数提高了0.1074。The paper discussed decision-tree classification principle and CART algorithm, introduced the texture information and NDVI index into decision-tree method to complete classification, and compared with the maximum likelihood classification results. Resuhs show that the Decision-tree classification method improves the overall ac- curacy of 8.9048%, Kappa coefficient of 0.1074 inceased by comparision with traditional classification methods relatively.

关 键 词：图像分类 决策树 纹理 NDVI CART算法 

分 类 号：P237[天文地球—摄影测量与遥感]