Building Information Modelling-based Smart Inspection Data Management for Unmanned Aerial Vehicle-enabled Visual Building Inspection

Abstract

Unmanned Aerial Vehicles (UAV) equipped with red, green, blue (RGB) and thermal cameras have been growingly used for visual building inspections. However, its full potential remains untapped due to challenges in managing fragmented and distorted images. While Building Information Modelling (BIM) has been envisaged as an effective platform for managing building lifecycle data, its potential for handling UAV inspection data has not been explored. To address this challenge, this doctoral research has investigated the use of BIM to establish a smarter and more efficient method for managing the UAV building inspection data. Firstly, a systematic review was conducted on the state-of-the-art UAV and BIM applications in visual building inspection. The review findings revealed four key research gaps. Secondly, the research conducted a comprehensive investigation in effective UAV thermal image acquisition strategy, focusing on three critical factors: temperature difference between building interior and exterior, ground sampling distance (GSD) of thermal images, and UAV oblique angle. The impact of these factors on inspecting façade anomalies was examined both qualitatively and quantitatively through laboratory and field experiments. Thirdly, the research proposed a BIM-based scheme for managing fragmented and distorted UAV images. An improved Generalised Hough Transform (GHT) method was developed for aligning RGB images with BIM by matching building façade features. However, this method faces certain obstacles when dealing with specific types of façades, such as curtain walls. To overcome this, a Structure from Motion (SfM) method was introduced as a supplementary solution. Additionally, perspective and geometry distortions on UAV images, especially when inspecting single-curved façades, were eliminated by BIM-based 3D surface unwrapping. Moreover, UAV thermal images were calibrated, corrected, enhanced, and registered onto BIM through multi-source image fusion and histogram-based correction. The effectiveness of the proposed scheme has been validated by computer simulations and field experiments, demonstrating its ability to convert fragmented and distorted UAV RGB and thermal images into a distortion-free panoramic image, seamlessly integrable into BIM. Finally, an implementation of a system that consolidates all developed approaches has validated the formulation of an effective UAV image acquisition strategy with efficient inspection data management in BIM.