基于遥感影像的国家级湿地保护区土地覆被类型信息提取  

Land cover type information extraction of national wetland reserve based on remote sensing image

作　　者：杨鑫[1] 于皓 刘长河 YANG Xin;YU Hao;LIU Chang-he(School of geomatics prospecing engineering,Jilin Jianzhu university,Changchun 130118,China;Huinan forest protection center of Jilin province,Tonghua 135102,Jilin province,China)

机构地区：[1]吉林建筑大学测绘与勘查工程学院,长春130118 [2]吉林省辉南国有林保护中心,吉林通化135102

出　　处：《吉林建筑大学学报》2023年第4期27-33,共7页Journal of Jilin Jianzhu University

基　　金：国家自然基金项目(93802502);第五批吉林省青年科技人才托举工程(93802501).

摘　　要：本文以哈尼、龙湾国家级湿地自然保护区为研究区、Sentinel-2多光谱卫星影像为基础数据源,采用面向对象分类方法、最大似然方法实现哈尼、龙湾国家级自然保护区土地覆被信息提取,在ArcGIS中创建随机样点并结合实地调查样本,验证两种分类结果的精度.结果表明,面向对象法比最大似然法分类结果更好,总体分类精度达到了95%,远大于最大似然法的77%,这是因为面向对象法消除了传统基于像素中同物异谱、同谱异物的影响,较传统的分类结果减少了许多错分现象,分类精度更高.Taken Hani and Longwan national wetland nature reserve as the research area and Sentinel-2 multi-spectral satellite image as the basic data source,the object-oriented classification method and the maximum likelihood method were adopted to extract land cover information in Hani and Longwan National Nature Reserve.Random sample points were created in ArcGIS and field investigation samples were combined to verify The precision ofthe two classification results.The results show that the object-oriented method is better than the maximum likelihood classification results,and the overall classification accuracy reaches 95%,which is much higher than the maximum likelihood method’s 77%.This is because the object-oriented method eliminates the influence of the traditional classification results based on the different spectrum of the same objects in pixels and the foreign objects in the same spectrum,and the classification accuracy is higher than the traditional classification results.

关 键 词：Sentinel-2 面向对象 ECOGNITION 土地覆被 

分 类 号：P237[天文地球—摄影测量与遥感]