Scan2Bim

Digital 3D models are key to increasing efficiency and economic viability in the construction industry. This applies to construction planning, renovation, construction progress monitoring, construction documentation and modular construction. Today, most new buildings are planned digitally, but for existing buildings, there usually is no digital data available.

Automated capture of existing buildings

Fraunhofer IPM offers an automated process for creating BIM-capable 3D models of existing buildings. The measurement data serves as a starting point: Powerful sensor systems capture interior spaces and façades. Our hand-held or robot-mounted mobile measurement systems consist of cameras, laser scanners and GNSS sensors that capture their surroundings quasi statically. For large areas, drone-borne sensors are used.

Object recognition with point clouds

In an AI-based process, 3D measurement data is automatically semantically enriched, instantiated and converted into a BIM-capable format – resulting in significant time savings over today’s standard manual model generation. Semantic segmentation, i.e. the detection of specific objects in the measurement data, is performed directly in the point cloud using a specially trained artificial neural network (ANN). In the past, typical object classes such as doors, windows, pipes and stairs were manually identified and entered as meta information into the data set. This AI-based approach is based on our extensive expertise in handling extremely large point clouds and in specific challenges such as reflections, transparency and edge detection.