A Knowledge Based Hierarchical Classification Tree for 3d Facade Modeling Using Terrestrial Laser Scanning Data

Due to high point density, accuracy and cost-effectiveness, Terrestrial Laser Scanner (TLS) do have the great advantage of providing the details of building facade as required for the production of realistic 3D city models. Thus, a significant amount of research efforts has been directed towards reconstructing building facade models using TLS data. Most of reconstruction strategies are based on geometric feature extraction, which is categorized into two groups: point-based and segmentation-based boundary extraction. However, little attention was paid on further recognizing and separating those extracted geometric features, result of which is helpful for 3D building modeling at LOD3. This paper presents an approach to recognize 3D facade features automatically from TSL data based on heuristic knowledge of geometrical and topological configuration of facade objects. All the facade knowledge is stored in a hierarchical tree, which guides the classification process. Finally, a confusion matrix is created to evaluate the classification performance.