Derivation of Roof Types by Cluster Analysis in Parameter Spaces of Airborne Laserscanner Point Clouds

The paper describes a method that extracts planar roof faces of building objects from point clouds obtained by a pre-segmentation of airborne laserscanner data. The method utilises a TIN-structure that is calculated into the point cloud. The parameters of every TINmesh are mapped into a triangle-mesh parameter space, which is then analysed by cluster analysis techniques. Both the utilisation of a 2-D parameter space (containing only triangle-mesh slope and orientation) and a 3-D parameter space (containing all three plane parameters) are described and tested. By cluster analysis and available knowledge on possible roof shapes, significant planes are derived from triangle parameter space. Results show that the use of a full 3-D plane parameter space is superior to the analysis of triangle-mesh slope and orientation only. Well-defined building roof planes can be extracted successfully, while disturbances such as dorms on buildings or geometric discrepancies in laserscanner data strip overlaps may significantly reduce the applicability of the technique.