基于决策树模型的海岸带分类方法研究  

Land Type Classification of Coastal Zone Based on Decision Tree Model

作　　者：何厚军[1] 王文[1] 刘学工[2] 

机构地区：[1]河海大学水文水资源与水利工程科学国家重点实验室,江苏南京210098 [2]黄河水利委员会信息中心,河南郑州450004

出　　处：《地理与地理信息科学》2008年第5期25-28,共4页Geography and Geo-Information Science

基　　金：河海大学水文水资源与水利工程科学国家重点实验室开放基金项目(2007490911);“908专项”江苏省海岸带调查项目

摘　　要：海岸带地物分布复杂,地物混淆常造成海岸带提取困难。该文以江苏省粉砂淤泥质海岸为研究对象,运用图像光谱特征、纹理特征并引入地学知识,构建研究区遥感图像分类决策树模型,并利用ETM+图像进行海岸地物分类研究。结果表明:采用的决策树模型可以较好地结合纹理信息和地学知识,解决遥感图像中复杂地物分类过程中的混淆现象,分类精度达89.26%,比最大似然法分类精度提高了15.19%。Coastal zone is characterized by the mixing of various land types. Classification confusion is a serious problem when processing the coastal image. In the present study, the ETM＋ image is used to classify the coast zone land types for the mud-silt coast zone of Jiangsu Province~ Decision tree model was developed based on spectral characteristics, texture characteristics and geo-knowledge for the study area. It is shown that, by building a classification decision tree with well combination of texture features and geo-knowledge for the study area, the problem of classification confusion for some land types can be well solved. Compared with conventional maximum likelihood method, the decision--tree method increases the classification accuracy by 15. 19% ,achieving an overall accuracy of 89.26%.

关 键 词：海岸带 遥感 决策树模型 纹理 地学知识 

分 类 号：TP75[自动化与计算机技术—检测技术与自动化装置]