基于决策树算法的遥感图像分类研究与实现

Research and Implementation of Classification in Remote Sensing Image Based on Decision Tree Algorithm

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,结合分层思想的树分类技术,对广泛用于数据挖掘模型中的CART决策树算法进行改进,提出了一种基于人机交互的决策树算法,将其应用到遥感图像自动分类中,具有很好的弹性和鲁棒性,且分类结构简单明了,达到了更好的分类效果。以VC++6.0作为开发工具,定义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 allusion to the traditional classification＇s shortcoming of low effect when dealing with remote sensing data that has complex spatial-character distributing, combining with tree-classification technology which using delaminating method, Classification and Regression Tree （CART） algorithm is improved, which is used largely in data mining. And an algorithm based on human-computer interaction is put forward, which is applied to remote sensing image classification with better flexibility and robust, also with simple and clear class structure, having achieved better classified effect. And a special data structure is designed to realize this classifying system in VC ＋ ＋. Practice show that this system has good stability, alternateness and strong practicability.