基于面向对象CART决策树的土地利用分类研究

Research on Land Use Classification Based on Object-oriented CART Decision Tree

面向对象的CART决策树分类方法可解决目前流行的监督分类、非监督分类以及模糊分类方法中“同物异谱、异物同谱”引发的漏分、错分问题。该方法融入了形状和纹理特征进行分类,同时运用二级分类体系解决了相似地物因光谱、纹理不同而导致的地物错分问题,分类效果较好。利用楚雄市鹿城镇2013年GF-1号遥感影像进行土地利用分类。结果表明:(1)基于光谱、形状和纹理信息选取的19个特征变量开展面向对象的CART决策树分类,总体精度可达90.22%,其中林地分类的效果最好;(2)二级分类体系解决了耕地、裸地因光谱、纹理特征多样而产生的地物错分问题,总体精度提高了7.06%,Kappa系数提高了8.17%。

Aiming at the missed and misclassified problems caused by“the same thing with different spectrums,different things with the same spectrum”in currently popular supervised classification,unsupervised classification and fuzzy classification methods,the object-oriented CART decision tree classification method can be used to solve this problem.This method incorporates shape and texture features for classification,and uses a secondary classification system to solve the misclassification problem of similar features due to different spectra and textures,in which the effect is better.In this paper,we used the GF-1 remote sensing images of Lucheng Town in Chuxiong City in 2013 to carry out land use classification.The results show that(1)the object-oriented CART decision tree classification based on 19 feature variables selected from spectrum,shape and texture information,with a total accuracy of 90.22%.Among them,the best effect belongs to the classification of forest.(2)The secondary classification system solves the misclassification problem of cultivated land and bare land caused by the diverse spectral and texture characteristics,whose total accuracy is increased by 7.06%,and the Kappa coefficient is increased by 8.17%.