多分类器融合与单分类器影像分类比较研究  

A study on the classification and comparison of the multi-classifier fusion and the single classifier

作　　者：牛明昂 王强[2] 崔希民[2] 赵康年[1] 柴鹏辉 

机构地区：[1]青海大学地质工程系,西宁810016 [2]中国矿业大学(北京)地球科学与测绘工程学院,北京100083

出　　处：《矿山测量》2016年第4期11-15,共5页Mine Surveying

基　　金：国家自然科学基金面上项目(51474217)

摘　　要：针对单一分类器的不足,文中采用最短距离分类器、马氏距分类器、K-均值分类器对多光谱遥感影像进行分类,并在测量级的融合方法下进行多种分类器融合分类实验,最后采用混淆矩阵进行分类结果精度评价。实验结果表明:多分类器融合的遥感影像分类方法在精度上高于单一分类器分类。Because of the limitations of the single classifier, the paper classifies the multi -spectral remote sensing image with the minimum distance classifier, Mahalanobis distance classifier and K -means classifier and makes the multiple classifier fusion experiment in the measurement level, based on which the paper makes precision accuracy of the experimental results with Confusion Matrix. It shows that the remote sensing classification method of multi - clas- sifier is more accurate than the single classifier.

关 键 词：多分类器融合 混淆矩阵 Kappa系数 精度 

分 类 号：TP751[自动化与计算机技术—检测技术与自动化装置]