Inhaltszusammenfassung

BIM is one of the strongest drivers of the digital transformation in the AEC industry during the last decades. Thanks to BIM, all disciplines are able to participate in digital planning and construction processes at the beginning and have access to central data in order to identify issues better before they occur. In addition to planning and construction, the management of existing buildings profits from digital BIM solutions, whereby FM has so far contributed sparsely due to the lack of avai...BIM is one of the strongest drivers of the digital transformation in the AEC industry during the last decades. Thanks to BIM, all disciplines are able to participate in digital planning and construction processes at the beginning and have access to central data in order to identify issues better before they occur. In addition to planning and construction, the management of existing buildings profits from digital BIM solutions, whereby FM has so far contributed sparsely due to the lack of available building models (as-built BIM). The process for 3D as-built modeling based on surveying technology is referred to Scan-to-BIM (S2B). According to the state-of-the-art technology, 3D point cloud data are obtained from TLS and serve both registered and manually pre-processed as information foundation for modeling an as-built BIM. The as-built modeling process lacks automation due to missing, obsolete and complex information about captured objects, relationships and attributes as well as customizable purposes for usage of the BIM. To achieve auto-mation, the literature suggests extending the time-consuming and highly manual S2B process with additional compo-nents like semantic segmentation, object recognition and geometry fitting algorithms. Published research affirms the successful use of automated approaches on highly simplified buildings but less claiming to represent the complex reality. As a consequence, there is a high demand for new research approaches that rethink the established S2B process chain. The latest research results by the authors indicate a sufficient point cloud quality of inexpensive consumer products, such as the Apple iPad Pro for as-built modeling of indoor scenes. Considering this finding and the demand to provide resilient approaches for an automatic S2B process, this paper deals with the development of an intelligent 3D data acquisition method by LiDAR-based consumer hardware. The presented 3D data capture application provides a detailed and semantically structured point cloud using artificial intelligence based on high-resolution depth data gathered by vision technology. This is done by capturing a depth map with the integrated device sensors which then can be classified simultaneously using Apple's AR framework, that is able to distinguish data between eight categories. At present, a prototype application is being developed that is capable of generating detailed geometric data, combining them with semantic attributes in real-time and supplying a structured 3D point cloud for the remaining process. A key feature of this novel application in academia is the simultaneous 3D data acquisition and classification performed by optical low-cost technology. Initial results highlight great capability in the field of indoor data capture for as-built documentation. By combining semantic segmentation and AR-assisted 3D data capture, the S2B approach becomes less complex and thus is now attractive to the FM and non-engineering industries.» weiterlesen» einklappen

Autoren

Klassifikation

DFG Fachgebiet:
Informatik

DDC Sachgruppe:
Informatik