基于决策树的高光谱遥感影像分类方法研究-《计算机技术与发展》

文章信息/Info

Title:: Research on Method of Hyperspectral Remote Sensing Image Classification Based on Decision Tree

Author(s):: HUA Ye; ZHANG Tao; XI Hou-wei; WANG Yu-fei; HUANG Xiu-li; China Electric Power Research Institute

Keywords:: binary decision tree; hyperspectral remote sensing image; classification; best threshold; automatic building

摘要:: 为了验证将决策树算法用于高光谱遥感影像分类的可行性,提出了一种二叉决策树自动构建算法用于高光谱遥感影像分类。通过对高光谱遥感影像进行现场采样、对样本进行统计和训练,生成了一棵二叉决策树,从决策树中提取出分类规则,并对高光谱遥感影像进行分类。生成的决策树简单明了,分类规则易于理解,分类效率和精度都比较高,实现了高光谱遥感影像从数据降维、样本选择、样本训练、决策树生成、影像分类的＂一体化＂和＂自动化

Abstract:: In order to validate the feasibility of using decision tree algorithm for hyperspectral remote sensing image classification,it proposes a method of building decision tree automatically for hyperspectral remote sensing image classification. Based on hyperspectral remote sensing image on-site sampling, sample statistics and training, generate a binary decision tree, extract classification rule from the decision tree and classify the hyperspectral remote sensing image. The whole tree is simple and the classification rules are easy to understand. Both classification efficiency and accuracy are satisfactory. The study makes it ＂integration＂ and ＂automation＂ to reduce the dimensionality of hyperspectral data, sample selection, sample traimng,decision tree generation and image classification