Computation of detailed 3D building models for the land registry from remote sensing data

• Applicant: Professor Dr. Steffen Goebbels
• Subject Area: Geodesy, Photogrammetry, Remote Sensing, Geoinformatics, Cartography
• Term: since 2025
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 565633716

Virtual 3D city models are used for simulations (e.g. noise maps, lighting models, solar potential analyses, flood maps, heat demand mapping), for urban and construction planning, for the visualization of statistical data and for city marketing. In Germany and internationally, 3D city models with a level of detail (LoD) of 2 according to the CityGML 2.0 specification are currently in use: The buildings in these models typically have simple roofs, composed of a few parameterized roof shapes, and walls, but no roof superstructures, windows, doors, balconies, etc. This level of detail is not sufficient for applications such as Building Information Modeling (BIM), escape route planning, energy efficiency studies, 3D printing, etc. Therefore, in a joint project with the partner City of Dortmund, the development of a workflow for the calculation of building models at LoD-3 level of detail is planned on the basis of preliminary work. This is to use existing, model-based LoD-2 models, cadastral data, point clouds and oblique aerial images. In particular, façade and roof objects such as doors, windows, balconies, dormers and chimneys are to be recognized in the geodata and automatically converted into model components. The workflow should be suitable for computing models of complete cities. Established quality criteria are to be adhered to. Based on the preliminary work, algorithms are planned that combine elementary computer vision techniques with deep neural networks. One research focus will be the development of suitable training data and strategies that take into account the resolution of the available data.

• DFG Programme: Research Grants (Transfer Project)
• Application Partner: Stadt Dortmund