Crack Identification for Bridge Structures Using an Unmanned Aerial Vehicle (UAV) Incorporating Image Geometric Correction

Abstract

Crack assessment of bridge structures is essential for maintaining safe transportation infrastructure. Traditional crack detection by manual visual observation has drawbacks, as it is expensive, time-consuming, and limited by the height and volume of bridges. Recently, unmanned aerial vehicles (UAVs) with image processing have been used to address these limitations. However, cameras on UAVs will generally not be perpendicular to the crack surface during actual measurements; therefore, deviation in the perspective angle can lead to inaccuracies in crack identification. In this work, we propose a robust and straightforward crack detection method based on geometric correction and calibration algorithms to address these issues. Four parallel laser emitters were installed on the UAV camera for crack image acquisition, and the laser-obtained images were geometrically adjusted using the four-point linear correction algorithm. After crack image processing, the object-to-image resolution was analyzed, and the crack information was extracted. Compared to the lens imaging concept method, the method based on the four-point lasers showed greater precision for crack width identification, with a measurement accuracy of over 95%. This indicated that the proposed crack identification system showed great potential for actual crack detection of bridges.