Evaluating Vision-Based Quantification Approach for Concrete Crack Inspection

Abstract

For a variety of uses, such as traffic tunnels, and utilities, concrete tunnel structures have been constructed for a very long time ago. Despite the restricted land area, it is required to build numerous infrastructures. As industries and technologies were developed, in order to solve this issue, underground spaces and tunnels for multiple purposes were constructed for efficient space usage. Numerous structures have degraded, and to systematically manage them, the inspector directly detects and evaluates cracks, leakages, and damage using cameras and measuring tools. However, this method is subjective and time-consuming. To overcome this, even though numerous studies utilizing computer vision are being conducted, the measurement error still high depending on the shooting condition. In this study, the performance of RGB-D camera-based measurement method was compared through laboratory-scale experiments. The proposed measurement method is significantly affected by the performance and working conditions of the segmentation algorithm. Therefore, the effect of each variable on the measurement result was compared and reviewed in terms of the field application of each method.