By taking QuickBird highresolution remote image of Chongqing University Town as data source, based on multiscale segmentation algorithm to achieve inner object with small difference, the typical land image feature database is constructed to implement the pertinent description of different land features, the cascade method is used to comprehend the advantage of nearest neighboring algorithm and fuzzy classification algorithm, the objectoriented extraction of mountainous city land cover information is implemented, and the extracted results are analyzed and compared with traditional monitoring classification method. The results show that the objectoriented classification of comprehensive utilization of spectral features, texture features and geometry features has higher classifying efficiency and stability, that this technology can effectively overcome the salt and pepper noise, that its classification result has good area feature and connection, that the classification accuracy of the objectoriented to arable land, grass land, construction land and road land with near or the same spectral features in mountainous city increases 28.67%, 9.24%, 10.38% and 8.79% respectively to effectively overcome the phenomenon of the same spectrum from different materials and the same materials with different spectrum. Comparison and analysis reveal that objectoriented classification method has good feasibility and certain application value in the mountainous city land cover information extraction, meanwhile, because of the complexity of the studied area and the universality of studied object, this algorithm research can be used in other mountainous cities. 