Vision-based measurement method with segmentation for concrete tunnel crack inspection

Abstract

Many underground infrastructures have been built as technology has advanced, and today, after 30 years or more, they are in a state of deterioration, necessitating safety management. As the necessity is emphasized, many studies using computer vision are being conducted to overcome problems such as the high cost and lack of objectivity of the conventional method that relies on manpower. Many studies are focused on deep learning-based crack detection, and research on crack quantification essential for inspection is still insufficient. In this study, a method to obtain the crack width of a concrete surface based on computer vision was proposed through a lab-scale experiment and compared with the previously studied method. Methods to increase applicability by overcoming the limitations of RGB-D cameras were additionally presented and analyzed. The results of this study are expected to play a significant role in the future development of automated systems for concrete tunnel cracks.