多分类器融合的遥感影像分类方法实验研究  

Design and implementation of remote sensing image classification program based on multi-classifier fusion

作　　者：孙皓 甘淑[1,2] 袁希平 SUN Hao;GAN Shu;YUAN Xiping(School of Land and Resources Engineering,Kunming University of Science and Technology,Kunming 650093,China;Yunnan Institute of Engineering Research and Application of Plateau Mountain Spatial Information Surveying and Mapping Technology,Kunming 650093,China;West Yunnan University of Applied Sciences,Dali 671000,China)

机构地区：[1]昆明理工大学国土资源工程学院,云南昆明650093 [2]云南省高校高原山区空间信息测绘技术应用工程研究中心,云南昆明650093 [3]滇西应用技术大学,云南大理671000

出　　处：《全球定位系统》2020年第5期108-112,共5页Gnss World of China

基　　金：国家自然科学基金(41561083,41861054);云南省自然科学基金(2015FA016)。

摘　　要：在遥感影像分类应用中,使用不同种类的分类器对原始图像进行操作所得出的分类结果的分类精度和效率都是不同的,设计完成多分类器融合分类实验,实验可以将不同种单一分类器的优点用适当的方法进行组合,从而获得比单一分类器分类精度和效率更好的分类方法.在此基础上采用投票原理设计抽象级融合的多分类器并完成实验.设计结果表明:多分类器融合的遥感影像分类方法在精度上要比单一分类器的分类精度高.In the application of remote sensing image classification,the classification accuracy and efficiency of the classification results obtained by using different kinds of classifiers to operate on the original image are different.In this paper the multi classifier fusion classification experiment are designed and completed.The experiment can combine the advantages of different single classifiers with appropriate methods,so as to obtain better classification accuracy and efficiency than single classifier Method.On this basis,the voting principle is used to design the multi classifier of abstract level fusion and complete the experiment.The design results show that the classification accuracy of multi classifier fusion is higher than that of single classifier.

关 键 词：遥感影像分类 单一分类器 多分类器融合 精度评价 

分 类 号：TP751[自动化与计算机技术—检测技术与自动化装置]