遥感图像决策树分类器研究与实现  

Research and Implement of Decision Tree Classifier in Remote Sensing Image Based on VC++

作　　者：罗来平[1] 宫辉力[1] 赵文吉[1] 刘先林[1] 

机构地区：[1]首都师范大学资源环境与地理信息系统北京市重点实验室&三维获取与应用教育部重点实验室,北京100037

出　　处：《遥感信息》2006年第3期13-15,i0001,共4页Remote Sensing Information

基　　金：国家863计划"遥感数据处理软件"课题(2003AA135010);北京市重点实验室开放基金项目共同资助

摘　　要：针对传统分类方法在处理空间特征分布极为复杂的数据时效果不佳的缺点[1],结合“分层思想”的树分类技术,将广泛用于数据挖掘模型中的CART决策树算法应用到遥感影像分类中,具有更好的弹性和鲁棒性,且分类结构简单明了,达到了更好地分类效果。并以VC++6.0作为开发工具,设计了一种特殊的数据结构,实现了该分类系统。实践表明,该系统具有很好的稳定性和交互性,实用性较强。In view of the low effect of the traditional classification when dealing with remote sensing data that has complex spatial distribution, combining with the tree-classification technology which adopts the layer concept, this paper applies the Classification and Regression Tree （CART） algorithm to the remote sensing image classification. It is robust and has better flexibility and simple and clear class structure, while it achieves better classification effect. A special data structure is designed to realize this classification system which developed by VC ＋＋ . Practice shows that this system has good stability, alternateness and strong practicability.

关 键 词：分类器 影像分类 决策树 遥感 VC++ 

分 类 号：P237.9[天文地球—摄影测量与遥感]