中文摘要：

CRUISE 2D决策树分类算法作为一种数据挖掘和知识发现的监督分类方法,综合了FACT,CART,QUEST决策树分类的思想.通过单因子和双因子交互检验和引导校正,快速有效地降低分割变量选择时产生的偏差,提高树的可读性,建立简单、高效、准确的决策树模型.基于CRUISE 2D决策树方法,以藏南地区为研究区,综合利用TM影像6个波段、NDVI,NDWI,SBI,GVI等波段信息,基于相同的训练样本和检验样本,利用判别规则建立决策树对影像进行分类;并将其与传统的监督分类方法 QUEST,SVM相比较,CRUISE 2D决策树分类方法总精度94.09%,比QUEST,SVM分类分别高10.86%,10.24%;Kappa系数0.931 0,比QUEST,SVM分类分别高出0.126 8,0.119 6.结果表明：CRUISE 2D能有效的改善传统监督分类中的错分漏分现象,在遥感分类上具有很高的稳健性和鲁棒性.

英文摘要：

Cruise 2D decision tree algorithm is one of supervised classification method in data mining and knowledge discovery. It is combined methods of FACT, CART, QUEST decision tree. Using one factor or two factor effects and a bootstrap adjustment prior to variable selection bias, CRUISE can improve the interpretability of its tree and make a easy, efficient and accurate model. The Landsat TM image was classified in Southern Tibet. And the CRUISE 2D decision tree precisely obtained new discriminant classification rules from integrated satellite image , NDVI , NDWI , SBI , GVI and other investigation information based on the same training and testing samples . Finally, the image was classified with the CRUISE 2D decision tree, and the result was compared with that of QUEST （Quick, Unbiased, and Efficient Statistical Tree） and SVM （Support Vector Machine） image classification . The overall accuracy was 94.09% , which was higher 10.86% ,10.24% than the overall accuracy of QUEST,SVM . Meanwhile , the Kappa efficient was 0. 931 0 , which was higher 0. 126 8,0. 119 6 than the Kappa efficient of QUEST, SVM . The results show that CRUISE 2D effectively improve the unclassified and misclassified pixels in the traditional supervised classification, and it has a very high robustness.