Taking Yangling demonstration zone of Shanxi province as the research area, using LANDSAT and DEM image data, the paper analyzed the spectral features and NDVI image characteristics of the study area, selected the appropriate classification method and established the right model of study area, classified the remote sensing image decision tree, selected 500 sampling points in map of gooleearth Yangling demonstration area. Through scene investigation, it took notes the data of selection point and counted the type of water body of statistical sampling points. The experiment indicated that the method of decision tree classification significantly increases classification accuracy and satisfies the need of production.