面向对象的密云县土地利用类型遥感分类方法

Research on Object Oriented Remote Sensing Classification of Land Use Types in Miyun County

为减弱遥感分类过程中出现的"椒盐现象",采用面向对象和决策树相结合的分类方法,在eCognition软件中对密云县的Landsat8遥感影像进行分类,并提取出了耕地、林地、建设用地、园地、未利用地和水体六大类型,期间为确保合理性又单独提取出阴影类。研究结果表明,分类后处理的影像总体精度和Kappa系数分别为88.2%和0.848,且该方法从一定程度上避免了分类结果中存在的"椒盐现象",为遥感影像的土地利用分类提供了参考。

In order to reduce the phenomenon of ＂salt and pepper＂ appearing in the classification of remote sensing,by using the classification method based on object-oriented and decision tree,Landsat8 remote sensing images in Miyun County were classified by eCognition software.The cultivated land,forest land,construction land,Garden,unused land and water body were extracted,and then the shadow category was also extracted to ensure the rationality.The results show that the overall accuracy and Kappa coefficient of the classified post-processing images are 88.2% and 0.848,respectively,and this method to some extent avoids the phenomenon of ＂salt and pepper＂ existing in the classification results,and it provides a reference for the land use classification of remote sensing images.