期刊名称：ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences
印刷版ISSN：2194-9042
电子版ISSN：2194-9050
出版年度：2008
卷号：XXXVII Part B7
页码：595-600
出版社：Copernicus Publications
摘要：The wetland is one of the most important ecological systems. The present condition and its change trend of wetland are very important to the making of policy upon the reclamation, exploitation, management and protecti on of wetland. In this paper, the coastal wetlands in the northern part of Jiangsu Province are taken as study object and the technology about the extraction of wetland is explored by using multi-features and knowledge rules and multi-spectral Landsat 7 ETM+ acquired on May 26, 2005, in combination with the analysis upon the characteristics of wetlands and its presentation in remotely sensed imagery and data of field investigation of the same period. Based on the analysis of the characteristics of spectrum about the wetland, in this paper, at first, unsupervised classification on the image of study area was conducted. And then, by using the spectral feature of wetlands, texture, principal component analysis, NDWI and relative knowledge rules, the results of unsupervised classification was improved. The accuracy of extraction of wetlands has been greatly improved, from 71.09% to 87.16% and KAPPA coefficient from 0.6546 to 0.8438. The results showed that the classification accuracy of the extraction of wetland using knowledge rules has a very great improvement. By the support of remote sensing software such as ERDAS, ArcGIS and S-PLUS, we have improved classification accuracy of wetland at 72.76%. By comparison, the classification accuracy of extraction by using knowledge rules on the results of unsupervised classification is very high
关键词：Land Use; Data Mining; Multispectral; Knowledge Base; Mapping; Landsat
