面向对象的黄河源典型区湿地信息提取  

Wetland information extraction for typical headwater region of Yellow River based on object-oriented approach

作　　者：安如[1] 陈志霞[1] 陆玲[1] 杨仁敏[1] 李晓雪[1] 吴红[1] 龚天宇[1] 曲春梅[1] 

机构地区：[1]河海大学地球科学与工程学院,江苏南京210098

出　　处：《河海大学学报（自然科学版）》2011年第4期355-360,共6页Journal of Hohai University(Natural Sciences)

基　　金："十一五"国家科技支撑计划(2006BAC08B00);国家重点基础研究发展计划(973计划)(2009CB421105)

摘　　要：以黄河源玛多县为实验区,遥感TM影像和DEM为数据源,在研究区湿地定义、成像机制及其地学规律分析的基础上确定了湿地遥感提取的分类体系和湿地信息提取的有效特征(包括光谱、纹理、地形和结构特征),并构建了面向对象的遥感湿地提取方法.实验显示:该方法湿地提取的总精度为90.13%,Kappa系数为88.33%,比常用的最大似然监督分类方法和决策树分类结果的精度有较大提高;采用该方法可以减少目视解译的工作量和主观性.Taking Maduo County,located in the headwater region of the Yellow River,as a study area,and TM remote sensing images and a DEM as test data,the classification system of wetlands were determined firstly based on the definition of wetlands,and the analysis of imaging mechanisms and geographical features.Then,the effective features for wetlands information extraction,including spectrum,texture,terrain and configuration,were studied.Finally,a method for wetland information extraction based on object-oriented technology was developed.The experimental results show that the method has a total precision of 90.13% and a Kappa coefficient of 88.33%.It greatly improves the precision compared with the maximum likelihood supervising classification method and decision tree classification method,and it also can reduce workload and subjectivity.

关 键 词：湿地 信息提取 面对对象 黄河源典型区 TM影像 

分 类 号：P467[天文地球—大气科学及气象学]