Abstract: As a kind of renewable resource, vegetation, including both natural and cultural ones, has long been a subject studied earnestly by remote sensing application specialists. Vegetation classification is one of the significant aspects of vegetation research. The use of satellite image to classify vegetation has been studied by important scholars both at home and abroad. This paper chooses the vegetation of the waters and the sand bars in the Dongting Lake as the subject matter for study, and tries to study the method of vegetation classification with remote sensing data. Based on the mutual dependent relationship between vegetation and environment and their comprehensive reflections in remote sensing images, the paper advances an analystic method of landscape-ecology analysis Jay these images, i. e. to combine the tone and shape structure features in the image with the ecological rules of the community to analyse, and consider this as the basis of the methodology for vegetation visual interpretation. On this basis, the waters and sand bars in the Dongting Lake are divided into 5 landscape-ecological models, and the vegetation of lake regions is divided into 17 basic vegetation types.