Automatisierte Generierung von 3D-Gebäudemodellen aus vektorbasierten Grundrissplänen

Floor plans are commonly used in the construction industry and contain crucial information about the depicted building. For several years now, the creation of such plans has heavily relied on computer systems, storing relevant data in vector file formats. However, due to the many different graphical representations of symbols and the lack of a unified drawing standard, the automatic interpretation of the semantic information in these floor plans is still an open problem. The main benefit of automated acquisition of this information lies in the possibility to efficiently generate three-dimensional building models, which can then be employed for different purposes. The goal of this work was to design and develop a system for the automated interpretation of vector-based floor plans. Further, the system should be able to make structural and topological information in form of three-dimensional building models available for interpretation by other applications. Next to the autonomous operation of the system, genericity and robustness were key factors and made up the main problem statement. Different floor plans have been analysed to establish a compact set of rules, which go well together with the developed recognition methods. Each method devised was enhanced iteratively until a satisfactory level of performance had been achieved. This resulted in a fully automated approach supported by various underlying semiautomatic methods. This approach provides a reasonable compromise between automation and genericity. The concluding quantitative evaluation of the developed methods yielded compelling results. The system was able to analyse complex floor plans with minimal user interaction in a reasonable time.