摘要
将多分类器集合应用于"北京一号"小卫星多光谱遥感数据土地覆盖分类,首先构建分类器集合,应用最小距离分类、最大似然分类、支持向量机(SVM)、BP神经网络、RBF神经网络和决策树等进行土地覆盖分类,然后利用Bagging、Boosting、投票法、证据理论和模糊积分法等分类器集成方法,得到综合不同分类器输出的最终分类结果。试验表明,多分类器集成能够有效提高"北京一号"小卫星土地覆盖分类的精度,具有广泛的应用前景。
Using the multi-spectral remote sensing image of Beijing-1 micro-satellite in Xuzhou city as the data source,it is shown that the multiple classifier combination can improve the classification accuracy effectively.Firstly,the different member classifiers,including Minimum Distance Classifier,Maximum Likelihood Classifier,Support Vector Machine,BP neural network,RBF neural network and decision tree,are used to land use/land cover classification.Then different multiple classifier integration strategies,including bagging,boosting,majority voting,Dempster-Shafer theory and fuzzy integral are used,to combine the results from the different classification method.Experiments show that the combination of multiple classifiers not only has the potential to improve the accuracy,but also has broad application prospects.
出处
《遥感信息》
CSCD
2011年第2期69-78,共10页
Remote Sensing Information
基金
国家高技术研究发展计划(863计划)专项课题(2007AA12Z162)
国家自然科学基金(40871195)
北京宇视蓝图信息技术有限公司北京一号小卫星应用开放基金联合资助
关键词
“北京一号”小卫星
多分类器集成
土地覆盖分类
Beijing-1 small satellite
multiple classifier combination
land cover classification
