TM影像决策树分类中的影响因素研究  

Factors Affecting Decision Tree Classification Method over TM Image

作　　者：张连华[1,2] 庞勇[1] 岳彩荣[2] 李增元[1] 范应龙[1] 谭炳香[1] 车学俭[1] 

机构地区：[1]中国林业科学研究院资源信息研究所,北京100091 [2]西南林业大学林学院,云南昆明650224

出　　处：《林业科学研究》2014年第1期1-5,共5页Forest Research

基　　金：亚太森林恢复与可持续管理网络项目"Forest Cover and Aboveground Biomass Mapping in the Greater Mekong Subregion and Malaysia"(编号:2011PA004);国家863课题"全球森林生物量和碳储量遥感估测关键技术(编号:2012AA12A306)"资助

摘　　要：以云南省西双版纳州一景TM影像为例,分析了影响分类回归树方法的主要因素。结果表明在其他因素均一致的情况下,训练数据如果使用涵盖各类别的外业调查数据比使用系统布设的训练数据分类精度更高,并且多种参数波段的选择也会有效地提高分类的精度。Taking one scene TM image of Xishuangbanna of Yunnan as an example, the main factors affecting the classification and regression tree method were analyzed. The results show that in the parameters under identical cir- cumstances, the training data has higher classification accuracy if the field investigation data covering all the classi- fication data were used rather than the system layout data. It also shows that selecting various bands of parameters can also improve the precision of classification effectively.

关 键 词：决策树分类 TM影像 训练数据 植被指数 波段组合 

分 类 号：TP751.2[自动化与计算机技术—检测技术与自动化装置]