Zusammenfassung
Summary
A novel approach is introduced which is based on the detection of buildings and building changes from high-resolution aerial images using artificial intelligence (AI). The data sets necessary for training the AI system are the building outlines from the official Digital Cadastre Map (DFK) and Digital Orthophotos without building lean (TrueDOP). The semantic detection of the buildings and building changes is carried out based on optical images and digital surface models, such as the normalized digital surface model (nDOM) and a Temporal Digital Surface Model (tDOM) derived from the difference of two time epochs.
The results of the current cooperation between the Bavarian Agency for Digitisation, High-Speed Internet and Surveying (BVV) and the Technical University of Munich (TUM) on the detection of buildings and building changes are presented and compared with results derived from existing approaches. The presented AI-based solution is basically transferable to all surveying administrations of the federal states nationwide.