The Study On Classification Technology Of Forest Vegetation In Beijing Mountain Area

Forest is an important coverage type of the earth,which plays an irreplaceable role in keeping balance of global eco-environment. The accurate extraction of forest vegetation information is one of the most important problem in forest resource investigation and management. Because of the effect of topography,information extraction of forest in mountain areas becomes more difficult.Therefore, searching for effective methods of information extraction is an urgent problem. In this study,the sites were located in northwest mountain area and Badaling forest farm in Beijing.,and images from Beijing-1 satellite and SPOT-5,together with the geographical auxiliary data were used to extract the vegetation features.By using fuzzy c-means clustering and support vector machine as the main classification methods, following aspects were concluded. Firstly,Gram-Schmidt was the best method of image fusion in moutain area. The second conclusion is that fuzzy c-means clustering can improve classification accuracy of forest vegetation in mountain area. In this study, it also showed that support vector machine was a effective classification method,which can improve classification accuray of forest vegetation, and texture features derived from gray level co-occurrence matrix integrated with spectral information can also improve classification accuracy. However, the size of texture windows affected the classification results. By comparison and analysis ,it showed that 3×3 and 5×5 windows were the best windows to extract the texture features, while the different nucleus functions had not obvious affection to the results...