In this study, a multilevel change detection method is proposed for buildings using airborne laser scanning point cloud data and GIS vector data. To detect detailed information of each changed building, the roof points and corresponding roof segments in laser scanning data are obtained by a multi-entity based classification method and a connected component analysis with defined rules based on prior knowledge. Then the alpha-shapes algorithm is performed to detect the edges of the roof. Finally, the proposed change detection method is used to detect changes between buildings in laser scanning data and GIS data. Experiment results suggest that both qualitative and quantitative analysis of changes can be obtained by this method, and the overall correctness of change detection achieves to 95%. Moreover, compared with most change detection methods which detect changes as demolished or newly built building, the method in this article can detect much more specific changes, the changed edges, for partly changed buildings.