像元信息分解和决策树相结合的影像分类方法

该文提出了一种基于像元信息分解和决策树相结合的遥感自动分类方法。选择广州市番禺区作为研究区。用像元信息分解和多变量决策树法把TM影像分为水体、植被、水泥地、土壤4种基本组分,分离成4类树枝;分别以4种基本地物组分作为分类掩膜,采用BP神经网络分类、形状指数提取、光谱特征提取等复合方法进行分枝,并开展野外遥感调查,以提高和验证分类精度。结果表明：该方法保证了分枝时地物的纯洁度,有效地避免了地物提取时多余信息的干扰和影响,提高了分类精度。结合实地调查数据与最大似然分类算法进行对比实验,表明该模型比最大似然总体分类精度高16％。

An Automatic Classification Model of Remote - Sensing Image Based on Pixel Information Decomposition Combined with Decision Tree

The authors propose a new automatic classification model of remote-sensing image using pixel information decomposition combined with decision tree in the study area of Panyu District,Guangzhou.At first,the Panyu TM image is divided into four elementary components that include water,vegetation,cement ground and soil by pixel information decomposition associated with multi-variable decision tree (four branches).Then,based on the elementary components,the authors continue to subdivide by BP neural network classification,shape index extraction and spectrum reflecting properties analysis.Finally,carry out field investigation to improve and validate the classification precision.The results show this method ensures the purity of branch objects and eliminates the disturbance and influence of unwanted objects effectively,so as to improve the classification precision.In comparison with the maximum likelihood classification by field survey data,the classification precision of this model heightens 16%.