Jiangsu Yancheng Coastal Wetland Classification Of Remote Sensing And Landscape Change Research

Remote sensing and GIS technologies are significant tools for data acquisition and quantitative analysis in the macro study of wetland. Furthermore, the remote sensing classification information of high precision for wetland directly affects the accuracy of wetland data, which is one of the important and difficult issues in remote sensing research of wetland.The core zone of Yancheng coastal wetlands are taken as a study area and the classification of wetland types is based on ALOS image in the thesis. Firstly, the traditional unsupervised classification has been improved. Secondly, modifying the unsupervised classification result according to the knowledge rules and GIS rules. This method is able to effectively recognize the wetland cover of ecotones and provide an ideal classification for wetland research of landscape-scale. The2006and2010remote sensing images are interpreted by the method above. So the dynamic evolution of the core area and the changes of the landscape can be illustrated. The main contents and conclusion are as follows:(1) Merging the initial unsupervised classification by using spectral angle of spectral matching techniques and spatial adjacency to avoid the inevitable confusions which are caused by similar spectrum. The errors caused by visual interpretation have been avoided, and thus it can provide a good start for subsequent image rectification.(2) We use the unsupervised classification method to conduct the primary classification system for the coastal wetlands. Then the reason for limiting the accuracy of wetland types were found, especially in the ecotones of wetland types. The ecotones among reed marsh, spartina alterniflora marsh and suaeda heteroptera kitag marsh have been revised by using the spectral feature of wetlands, texture, principal component analysis and relative knowledge. In addition, the rest parts were revised by GIS rules. Finally, the precision of classification was tested by the GPS data and the average accuracy of wetland types has reached90%, which suggested that the multi-level classification methods including the knowledge rules and GIS rules are effective in extracting the coastal wetland cover information.(3) By the way of the landscape change research of the core area, we found that the areas of reed marsh and spartina alterniflora marsh have been increased while the pond and suaeda heteroptera kitag marsh are the opposite Landscape diversity and evenness have been increased to some extent.