The Research On Extraction Methods Of Wetlands From Remote Sensing Image

Wetland, formed by the amphibious interaction and with special function, is an important part of the ecological system on the earth surface. Owning to the ignorance of people on wetland's ecological value, the wetland resources was badly damaged. At resent years, most of researchers come to the wetland research yield, so the Wetland research fields extend gradually from the pure theoretical research to multiple directions. Along with the gradual deepening research of wetland, the traditional methods on wetland survey can not meet the demand of information of wetland. Remote sensing technique has become an important data source of wetland research. By the same time, the extraction of wetland information from remote sensing images and analyses the wetland information has become an important yield of wetland research. However, because the spectrum of remote sensing image is the mixed one of various component values and the wetland has Great complexity, although many scholars made great efforts, the improvement on the wetland's accuracy of remote sensing classification still meets many difficulties. So this paper make a try to extract wetland from remote sensing image using decision tree:taking Ruoergai wetland at Sichuan province as the research area, the author utilizes the TM image obtained on August 5th, 2006 to test.Use the LUCC data by Institute of Geographic Science and Natural Resources research, the wetland data 2008 provided by Institute of Remote Sensing Application Technique China Academy of Sciences and STRM DEM information etc data as auxiliary information. Then, classification was made by employing the ISODATA, maximum likelihood, and neural network, support vector machine (SVM) and decision tree. After the classification,accuracy evaluation is processed by confusion matrix using remote sensing image processing software ENVI. Then analyze the precision of wetland categories. Experiment results indicate that the decision tree method decision tree classification method can obtain great accuracy of classification. At the same time, the cooperation results among the results of the different classification show that the decision tree classification result has the best accuracy.