Automated Damage Localization and Quantification in Concrete Bridges Using Point Cloud-Based Surface-Fitting Strategy

Abstract

Digital image processing is considered an alternative to manual visual inspection, enabling automated damage evaluation for structural maintenance. Although advancements in artificial intelligence have improved identification performance, directly quantifying the surface damage in three-dimensional (3D) space using only two-dimensional (2D) images is difficult. In addition, because close-up images are preferred owing to the high measurement accuracy, its application requires a considerable amount of time to process numerous images of full-scale structure. In this study, a framework for automated damage evaluation using 3D laser scanning is presented. The proposed approach is designed to process the point clouds of a full-scale bridge by addressing different shapes. Furthermore, a tailored fitting strategy is employed to accurately identify the surface damage on the edge, which can cause false detections. In practice, the performance of the proposed framework is systematically validated on the point clouds of the bridge components.