融合面向对象与缨帽变换的湿地覆被类别遥感提取方法  

Method for wetland type extraction using remote sensing combing object-oriented and tasseled cap transformation

作　　者：罗开盛[1,2] 陶福禄[1] 

机构地区：[1]中国科学院地理科学与资源研究所陆地表层格局与模拟院重点实验室,北京100101 [2]中国科学院大学,北京100049

出　　处：《农业工程学报》2017年第3期198-203,共6页Transactions of the Chinese Society of Agricultural Engineering

基　　金：国家自然科学基金(91325302)

摘　　要：为了有效提取湿地覆被类别遥感信息,该文基于国产环境星影像(HJ-CCD)和Landsat7遥感影像(ETM)提出了一种融合面向对象技术和缨帽变换的提取湿地覆被信息的方法,并对东洞庭湖区的湿地进行提取。遥感提取结果的总体精度90.02%,Kappa系数0.88,高于传统的分类方法分类的量化结果;获得的结果没有"椒盐现象"且比较紧致。试验结果表明融合面向对象和缨帽变换的方法能够有效的提取湿地覆被类别,精度高,效果好。研究结果为有效地利用遥感手段提取湿地覆被信息提供参考。Wetland is one of the most important ecosystems, and it has high social benefit, economic benefit and scientificresearch value. However, wetland resources are bearing a heavy pressure because of various natural and anthropogenic factors.The degradation of the wetland quality and quantity has aroused widespread concerns. To conserve and manage wetlandresources, it is important to monitor wetlands and their adjacent uplands. Satellite remote sensing has several advantages, suchas wild coverage, saving time and labor, multi-temporal, multi-platform, containing a large amount of information, and so on,when monitoring wetland resources especially in large geographic areas. In early work, the satellite imagery used the visualinterpretation for classification, which is still used widely today. The most commonly used computer classification methods areunsupervised classification and supervised classification. However, it is difficult to make great progress on improving theaccuracy of remote sensing classification because of ＂different things with the same spectrums＂ in wetlands. Spectrumconfusion among wetlands seriously restricts the extraction of wetland information and the application of remote sensingtechnology in the monitoring of the wetland. But the traditional pixel-based methods cannot overcome this difficulty because itonly used the spectral features of imagery, ignoring other information that the remote sensing imagery carries, although it hasbeen universally applied in land cover information extraction for many years. In order to over this difficulty and promote theapplication of remote sensing technology in dynamic monitoring of wetland, a new hybrid classification approach for wetlandwas proposed in this paper, which combined the object-oriented technology and the tasseled cap transformation method. Thenew proposed approach was further checked by a case study of wetland extraction based on the HJ-CCD and Landsat ETM（enhanced thematic mapper）remote sensing images in 2010 in the eastern Dongting Lake

关 键 词：遥感 分类 湿地 面向对象 HJ-CCD影像 缨帽变换 

分 类 号：P954[天文地球—自然地理学]