基于决策树的多源遥感数据分类

Multi-source Remote Sensing Data Classification Method Based on Decision Tree

遥感影像是获取土地覆盖信息的重要手段。分析了影响决策树分类的特征因子,并根据这些因子构建分类决策树。结合中分辨率多源遥感数据,对贡嘎山区进行土地覆盖分类,通过与最大似然法分析对比,基于决策树的多源数据分类对试验区的分类精度(总体精度85.71%,Kappa系数0.83)远高于基于像素的最大似然法监督分类(总体精度63.56%,Kappa系数0.58)。

Since it is an important approach to get land cover information,remote sensing provides services to resource surveys,environmental monitoring,etc.,the study of remote sensing image classification is significant. This paper analyzes the factors affecting the characteristics of decision tree classification,and then the decision tree to classify the image was built based on these factors. Combined with medium resolution multi- source remote sensing data,taking Gongga Mountain for instance,the comparisons to the maximum likelihood method were performed for the validation. The result demonstrates that the classification accuracy of the test area（ overall accuracy 85. 71%,Kappa coefficient of 0. 83） is much higher than the pixel- based maximum likelihood classification（ overall accuracy of 63. 56%,kappa coefficient of0. 58）,showing the advantages and prospects of the object- based multi- source data decision tree classification.