基于决策树方法的遥感影像分类

The Classification of Remote Sensing Image Based on Decision Tree

以齐齐哈尔市辖区为研究区域,利用分类回归树(Classification and Regression Tree,CART)算法从训练样本数据集中挖掘分类规则,集成遥感影像的光谱特征、纹理特征和地学辅助数据建立研究区的决策树模型.用实测的GPS样本点对分类结果进行精度验证,并与最大似然监督分类方法(Maximum Likelihood Classification,MLC)进行对比.结果表明,基于CART的决策树分类结果的总精度和Kappa系数分别为82.24%和0.77,分类精度较MLC监督分类方法有明显提高,有较好的分类效果.

In this paper, Landsat TM images of Qiqihar city in Heilongjiang were classified with a decision tree, which was established based on the analysis of the spectrum features, and other auxiliary information, such as NDVI and topography characteristics. Classification and Regression Trees （CART） algorithm was used for mining classification rules from the training sample data sets. Then decision tree classification with maximum likelihood classification was compared. The result indicated that the accuracy of decision tree classification was better than that of the maximum likelihood classification.