Typical airborne LiDAR data processing consists of ground/non-ground filtering, object classification and grouping, surface patch segmentation of the object, and modeling. This paper presents methods for building grouping to group individual building and roof patch segmentation. The methods are based on 3D adjacency among points and geometric characteristics of the surfaces. The proposed methods were applied to simulation and real LiDAR data. The results show that the point clouds were reasonably segmented both data.