Point Cloud Capturing and AI-based Classification for as-built BIM using Augmented Reality

Abstract

The benefits of using Building Information Modeling (BIM) have been proven in architecture, engineering and construction industry. However, implementing BIM in in facility management has not been achieved yet due to missing complete and accurate asbuilt BIM. Modeling comprehensive information for as-built documentation from 3D point cloud data is referred as Scan-to-BIM but lacks automation caused by unstructured data and high user input. We tackle the main issue of structuring the 3D point cloud data by using artificial intelligence while capture. With both, a highly reliable and low-cost technology we achieve less time-consuming point cloud capturing and segmentation contributing to a novel Scan-to-BIM approach with promising initial results.