Generalization of 3d Building Data Based on Scale-spaces

For an interactive visualization of three dimensional (3D) city models a Level of Detail representation is extremely useful to avoid unnecessary computations. Differently detailed models of the same object are displayed depending on the distance of the object. This implies that less detailed models have to be generated from complex models (generalization). In this paper an approach for the automation of 3D generalization especially for building data is presented. It uses the formally well-defined scale-space theory which includes morphology and curvature space. Practical investigations for the scale-space operations making use of the geometry kernel ACIS 3D Geometric Modeler show good results for morphology. For curvature space important preparations concerning the distinction between convex and concave object parts have been carried out. Tasks still to be performed include the coherent modeling and the implementation of continuous curvature space, the combination of morphology and curvature space, and the rectification of angles.