A New Method of Crack Width Estimation for Concrete Structures Monitoring

Abstract

In both aerospace and civil structures, cracks are considered critical defects which might cause significant damage if ignored. Currently, the most common method of crack inspection remains the visual one. This process is time-consuming and non-efficient. A cost-effective alternative is to use an automated system that could perform real-time crack detection and evaluation. To achieve this goal, many image-based crack detection methods were developed. However, the estimation of crack width, a significant factor in evaluating the structural condition, has yet to be explored sufficiently. The thickness of the cracks needs to be accurately quantified. In this paper, a crack width estimation approach based on computer vision is presented and evaluated. The proposed method in this study is based on segmentation, binarization and pixel-to-millimeter correspondence for detecting cracks and quantifying their thickness. It enables a robotic inspection system to evaluate images captured at any position and illumination conditions. The evaluation is achieved by using our own collected and annotated dataset.