作　　者： ; ; ; ;

机构地区： 中国矿业大学环境与测绘学院国土环境与灾害监测国家测绘局重点实验室

出　　处： 《中国矿业大学学报》 2011年第2期273-278,共6页

摘　　要： 针对不同分类器在遥感影像分类中的应用效果,将模式识别领域的研究热点———多分类器集成,引入煤矿区土地利用遥感分类.分别以国外Landsat ETM+和国产中巴地球资源卫星(CBERS)影像为数据源,按照煤矿区土地利用分类的特点和需求,构建由支持向量机、径向基神经网络、最大似然分类器、最小距离(马氏距离)分类器、J48决策树等组成的分类器集合,基于Double Fault,WCEC,Kappa等差异性测量指标选择成员分类器,利用Bagging,Boosting、加权投票法、分类器动态选择法、分层组合分类器等分类器集成方法实现组合成员分类器输出,获得集成不同分类器优势的分类结果.试验表明:多分类器集成能够有效地提高土地利用分类精度,在煤矿区土地动态监测和生态环境分析领域具有广泛应用前景. According to the effectiveness of various remote sensing classifiers to land use classification in coal mining area,multiple classifier combination is introduced to land use classification in coal mining area by remote sensing images.Using Landsat ETM＋ and CBERS remotely sensed images as data sources,five member classifiers,including support vector machine,RBF neural network,maximum likelihood classifier,minimum distance classifier and J84 decision tree,are used to classify land use at first and form the classifier pool,and then the member classifiers are selected based on classifier diversity measurement.Boosting,Bagging,weighted voting,classifier dynamic selection and hierarchical combination classifiers are used to integrate the outputs of member classifiers to obtain final results.The experiment results indicate that the multiple classifier combination can improve the accuracy of remote sensing classification for land use in coal mining areas as a promising strategy.

关 键 词： 煤矿区 土地利用分类 遥感 多分类器集成

领　　域： [自动化与计算机技术] [自动化与计算机技术]