Markerless Augmented Reality For Facility Management: Automated Spatial Registration Based On Style Transfer Generative Network

Abstract

Abstract

Access to on-demand and real-time building information is essential for enhancing efficiency in facility management. By integrating Building Information Modeling (BIM) with augmented reality (AR), facility managers can visualize and interact with digital building data directly in the physical environment. Despite advancements, the automated spatial registration between BIM and the real-world view remains a major hurdle. This paper introduces a novel markerless AR registration method powered by artificial intelligence and deep learning, specifically using a style transfer generative network. The proposed solution eliminates the need for manual alignment and enables seamless synchronization between BIM architecture and real-time camera input from mobile devices.

A mobile AR application was developed to implement and validate this approach. Results from field testing confirm the system’s capability to align BIM models with actual building views, allowing users to explore detailed digital overlays, including hidden infrastructure like mechanical, electrical, and plumbing (MEP) systems. This approach represents a significant step forward in applying BIM in the construction industry, improving the use of BIM modeling and BIM software architecture for smarter and more intuitive facility management.