Topological Reconstruction of 3D City Models with preservation of semantics

3D city models are becoming increasingly important for applications such as evacuation scenarios and energy consumption estimation. For these applications embedding semantic information on geometry is a key factor. The most popular implementation of modern 3D city models is based on the CityGML data model which describes spatial 3D data using a geometrical representation according to the GML encoding standard. While CityGML supports some basic storage of topological relationships between geometric objects, it fails to offer a true 3D topological representation of the city model. Alternatively, a true topological data structure can be used as an intermediate data model, to enable enforcing certain restrictions and operations that are more efficient for specific applications. In this article, we discuss a method that we have developed for the automatic conversion of CityGML models to a topological structure, while maintaining semantic information that was initially attached to the city objects. Such an approach raises certain challenges, as the geometries are not one-to-one analogous to the topological objects that are needed to represent them. We also provide a few examples that indicate that such a method is not trivial for retaining all information that was initially stored in a city model.