A concrete bridge crack size measurement procedure based on unmanned aerial vehicle and image registration

As cracks play a vital role for the durability of concrete bridges, crack detection becomes a necessity for concrete bridge inspection. However, the widely-used visual inspection for the crack detection is not only unsafe for inspectors, but also time and labor consuming. Thus, in this research, we propose a workflow that utilizes high-resolution digital cameras on unmanned aerial vehicles (UAVs) to measure concrete crack sizes. To be specific, we first use long-range images to establish the absolute coordinate system with control points and then register long-range and close-up images with the SIFT (Scale-Invariant Feature Transform). In this case, when measuring crack sizes on close-up images, we can derive the absolute size of each crack. Finally, crack sizes are measured manually on close-up images. An experiment was conducted on a concrete prototype in a laboratory. Comparing to the in-situ measurements from two surveyors, the proposed solution can measure crack lengths with 3.97mm root mean square error (RMSE), -0.93mm mean absolute error (MAE) and 3.58% mean relative error (MRE) and measure crack widths with 0.33mm RMSE, -0.0055mm MAE and 20.29% MRE. Overall, there are four main contributions in this preliminary research. (1) We propose a safe and efficient concrete crack inspection method by using UAVs. (2) We present a camera calibration workflow for digital zoom cameras to reduce errors caused by lens distortions. (3) For the long-range and close-up image registration, we demonstrate that the SIFT is an image matching method that can accommodate scale and orientation changes. (4) We identify potential issues for crack measurement that could help lead to future research improvements.