Markerless Augmented Reality for Facility Management: Automated Spatial Registration based on Style Transfer Generative Network

Abstract

On-demand and real-time building information is of great value to support facility management. Such information can be easily retrieved from an up-to-date building information model (BIM), and then intuitively presented to facility managers or inspectors by augmented reality (AR). However, effective spatial registration into BIM so as to align the virtual and real content still remains an unresolved challenge. Leveraging recent development in the field of generative networks, this paper proposes a markerless registration approach that can automatically align BIM with the real view captured by a mobile device without any manual operation. A mobile AR application is develop based on the proposed registration approach. Our field experiments demonstrate the effectiveness of the proposed approach for automated BIM registration. The successful registration thus allows users to access the rich building information, especially invisible utility such as the mechanical, electrical and plumbing (MEP) system, in the real-life context for better facility management practice.