Automated UAV image-to-BIM registration for building façade inspection using improved generalised Hough transform

Abstract

While Unmanned Aerial Vehicles (UAVs) are increasingly used for building façade inspections, their advantages have not been fully leveraged due to the challenge in gaining a fast and holistic understanding of as-is building conditions from fragmented images. This paper describes a framework to address the problem by registering UAV images into a Building Information Model (BIM) in three phases: (1) Extract position and optical parameters from UAV images to configure virtual cameras in BIM and generate template images; (2) Employ an improved generalised Hough transform (GHT) to extract building façade components with arbitrary shapes; (3) Project UAV images onto an orthophoto and incorporate them into 2D and 3D views. Finally, a Dynamo prototype is developed to automate the process. Computer simulations and field experiments confirmed a mean image-to-BIM registration error of within 21 mm, demonstrating the proposed approach has the potential to facilitate UAV-enabled building façade inspection.