3D building modeling can be more easily generated with several methods. LIDAR data have high potential of automation; though they have such disadvantages as low resolution and no texture. On the other hand, generally, a resolution of the aerial images is higher than that of LIDAR data. The extraction of buildings from aerial images in dense urban environment is easier from aerial images than from LIDAR data. However, the fully automated 3D modeling of buildings from aerial images in dense urban environment is not successful. The objective of this research is to develop a methodology to automatically generate 3D modeling of buildings. The generation of 3D buildings data is done by integrating high resolution air-borne linear CCD, low resolution DSMs (created from LIDAR) and existing 2D digital maps. The detail representation of roof shapes such as gables is created by using the knowledge of buildings.