Extraction and reconstruction of curved surface buildings by contour clustering using airborne LiDAR data

Airborne LiDAR (Light Detection and Ranging) can acquire detailed geometric information of topography and objects. In this paper, an automatic method for extraction and reconstruction of buildings with curved roof was investigated using airborne LiDAR data. Based on the distinct topological relationship and characteristics of contour lines, this method firstly clustered and derives the buildings’ contour lines automatically with several attributes such as the height value of contour line, the enclosed contour line's area and circularity. LiDAR points corresponding to different parts of the roof were separated according to the inter-topology of building contour lines. Then roof top type was identified through analyzing the relationship between the contour lines’ height values and their enclosed areas. Finally the roof top was reconstructed by fitting the corresponding separated LiDAR points according to its type, which especially took the continuity and smoothness of the connected curves into considering. To validate this method, an experiment was conducted on a small urban part with several different types of curved roof. The derived building models with curved roof demonstrated the proposed method's plausibility and this new curved surface approach could be used as a complemented method to existing planar surface method for automatically building reconstruction.