基于决策树分类算法的遥感影像土地利用分类

Land Classification of Remote Sensing Image Based on Decision Tree Classification Algorithm

基于专家知识的决策树分类算法是遥感分类技术中一种重要的技术手段.以内蒙古太仆寺旗为研究区,以Landsat8遥感影像为数据源,通过提取NDVI、NDWI、NDBI和纹理特征变量,利用决策树分类算法中的C5.0算法和CRAT算法,对比传统的监督分类方法,对影像进行土地利用分类,并对分类结果进行精度评价.结果表明:决策树分类法分类精度要高于传统的最大似然分类法,且C5.0决策树分类算法在影像分类结果精度上要优于CRAT决策树分类算法.

The decision tree classification algorithm based on expert knowledge is a kind of important technical means during the classification of remote sensing technology.Using Taipus banner in Inner Mongolia as the study area,Landsat8 remote sensing image as data source,through the extraction variables of NDVI,NDWI,NDBI and texture feature,C5.0 algorithm of decision tree classification algorithm and CRAT algorithm were classified the land use of image,compared to the traditional supervised classification method,and verify the accuracy of the classification result.Results show that the classification of decision tree classification accuracy is higher than the traditional maximum likelihood classification method,C5.0 decision tree classification algorithm in image classification result accuracy is better than CRAT decision tree classification algorithm.