The requirement for establishing the 3D geometric shape of urban objects is growing, not only for the purpose of urban visualisation, but also for the purpose of city design, planning and management. The former of these needs requires the 3D shape of urban objects with the objective to render these shapes with either typical or actual textures. In this application the need for geometric accuracy is not as stringent as in the latter application, with the main emphasis being visual aesthetics. However, as visualization becomes increasing more like “reality”, analysts must pay further attention to both geometric detail and geometric correctness. This paper discusses the methods and results of attempting to extract the exterior 3D geometry of buildings from airborne laser scanner data. We measured the “true” exterior shape of one multi-story building using terrestrial surveying methods. A number of techniques were utilized to extract the shape of this and other buildings from the ALS data. We used the software LIDAR Analyst and found that it was able, using rules and a neural learning approach, to determine roof shapes. Coordinates of critical points on the building, as extracted from the roof shape, were compared with coordinates from terrestrial survey. Results showed that a systematic planimetric error existing in the roof shapes. After removal of this systematic error, random errors were approximately 1m in X,Y and 0.5m in Z. 3D wire frames were exported from Lidar Analyst to a CAD package for rendering and imported back into other programs for visualisation.