基于决策树分类技术的遥感影像分类方法研究  

Classification Methods of Remote Sensing Image Based on Decision Tree Technologies

作　　者：申文明[1] 王文杰[2] 罗海江[2] 张峰[2] 刘小曼[2] 熊文成[2] 

机构地区：[1]中国科学院地理科学与资源研究所 [2]中国环境监测总站,北京100029

出　　处：《遥感技术与应用》2007年第3期333-338,共6页Remote Sensing Technology and Application

基　　金：科技部十五攻关项目资助(2003BA614A-06-04)

摘　　要：以河北唐山为研究区,应用Landsat ETM+影像数据和GIS数据,对决策树分类技术和传统计算机自动分类方法进行了比较。研究表明:决策树与传统自动分类方法相比,分类精度提高了18.29%,Kappa系数提高0.1878。在地形起伏的山区,应用DEM及其衍生数据等GIS数据作为辅助数据可以提高分类精度19.52%,Kappa系数提高0.281;反射率影像分类效果比原始DN值影像的分类效果好,分类精度提高15.86%;缨帽变换在压缩数据量的同时,分类精度有所降低。In this article, the author compared decisin tree classification technology with classic automaticclassification technologies using Landsat ETM ＋ image data and GIS data of Tangshan City in Hebei. The result of this research showed： accuracy of decision tree classification compared with classic automatic classification technology was improved about 18. 29%, Kappa coefficient was increased by 0. 1878; classification accuracy was improved about 19. 52% when DEM and its derivative data were used as ancillary data in the mountainous area, Kappa coefficient was increased by 0. 281~ the classification accuracy was improved by 15.86% when the DN （Digital Number） values were converted to at-satellite reflectance values; tasseled cap transformation could cause classification accuracy to be reduced appreciably accompanied by compression of data amount.

关 键 词：遥感影像 决策树 计算机自动分类 空间数据挖掘 土地利用/土地覆盖 

分 类 号：TP751[自动化与计算机技术—检测技术与自动化装置]