Abstract: Based on the Environment for Visualizing Images (ENVI) EX, the SPOT5 high resolution remote sensing images which are extracted from Google Earth are used to investigate the classification of land use types by object- oriented multi- scale image segmentation method in the Qionghai lake basion. The results showed that the classification method can overcome the problem of the "salt and pepper effect" by the traditional pixel- based classification, it also can make a effective use of the auxiliary information (DEM, NDVI, etc.) and the information of ground objects (spectral characteristics, texture features, etc.) comprehensively. The classification accuracy reached 89. 18% and the results are satisfactory.