基于多分类器集成的GF-2遥感影像分类研究  

Research on GF-2 Remote Sensing Image Classification based on Multi Classifier Integration

作　　者：龙飞 任金铜 Long Fei;Ren Jintong(Guizhou University of Engineering Science,Bijie,China;Guizhou Province Key Laboratory of Ecological Protection and Restoration of Typical Plateau Wetlands,Bijie,China)

机构地区：[1]贵州工程应用技术学院,贵州毕节 [2]贵州省典型高原湿地生态保护与修复重点实验室,贵州毕节

出　　处：《科学技术创新》2023年第3期19-24,共6页Scientific and Technological Innovation

基　　金：国家级大学生创新创业训练计划项目(项目编号:202010668075);贵州省教育厅青年科技人才成长项目(黔教合KY字[2020]149);毕节市科技计划联合基金项目(毕科联合字G[2019]15号);贵州省典型高原湿地保护与修复重点实验室开放基金项目;贵州工程应用技术学院2022年学科建设项目;毕节市六冲河流域生物保护与生态修复人才团队(编号:毕人领通(2021)12号);智慧地理空间信息应用工程中心资助。

摘　　要：影像分类是遥感信息提取的关键技术之一。针对传统单分类器分类精度低的问题,以夹岩水利枢纽工程项目建设核心区为研究区域,提出了一种面向高分二号(GF-2)遥感影像的多分类器集成分类方法,对工程项目建设区进行分类研究,并对比分析传统分类器的分类精度。结果表明:马氏距离分类法的总体分类精度为91.7%,Kappa系数为0.84,对水体分类效果较好;最大似然分类法的总体分类精度为95.7%,Kappa系数为0.92,对林地、水体分类效果较好,道路分类效果不够理想;支持向量机分类法的总体分类精度为96.60%,Kappa系数为0.94,对林地、水体、道路分类效果较好;多分类器集成分类方法的总体分类精度为97.80%,Kappa系数为0.96,其分类精度高于传统单分类器的分类精度。Absrtact:Image classification is one of the key technologies of remote sensing information extraction.Aiming at the problem of low classification accuracy of traditional single classifier, taking the core area of Jiayan Water Conservancy Project as the research area, proposed an integrated classification method of multiple classifiers oriented to the remote sensing image of GF-2, conducted classification research on the construction area of the project, and compared and analyzed the classification accuracy of traditional classifiers. The results showed that the overall classification accuracy of Markov distance classification was91.7%, and the Kappa coefficient was 0.84, good effect on water body classification;The overall classification accuracy of the maximum likelihood classification method was 95.7%, and the Kappa coefficient was 0.92,which was good for forest land and water body classification, but not ideal for road classification;The overall classification accuracy of the support vector machine classification method was 96.60%, and the Kappa coefficient was 0.94, which was good for forest land, water body and road classification;The overall classification accuracy of the multi classifier ensemble classification method was 97.80%, and the Kappa coefficient was 0.96, which was higher than the classification accuracy of the traditional single classifier.

关 键 词：多分类器集成 高分二号 夹岩水利枢纽工程 

分 类 号：P237[天文地球—摄影测量与遥感]