Novel visual crack width measurement based on backbone double-scale features for improved detection automation

• Published in: Engineering structures Vol. 274; p. 115158
• Main Authors: Tang, Yunchao, Huang, Zhaofeng, Chen, Zheng, Chen, Mingyou, Zhou, Hao, Zhang, Hexin, Sun, Junbo
• Format: Journal Article
• Language: English
• Published: Kidlington Elsevier Ltd 01-01-2023; Elsevier BV
• Subjects: Algorithms; Automation; Concrete crack; Image thinning; Machine vision; Multi-scale feature fusion; Vision systems; Image thinning; Concrete crack; Machine vision; Multi-scale feature fusion
• ISSN: 0141-0296; 1873-7323
• DOI: 10.1016/j.engstruct.2022.115158

•Backbone neighborhood distribution points are reduced to facilitate classification.•Dual-scale backbone features are combined for accurate width measurement direction.•A detailed visual measurement process of crack width is proposed.•A visual measurement method of crack width that is closer to reality is used to obtain more accurate results.
State-of-the-art machine-vision systems have limitations associated with crack width measurements. The sample points used to describe the crack width are often subjectively defined by experimenters, which obscures the crack width ground truth. Consequently, in most related studies, the uncontrollable system errors of vision modules result in unsatisfactory measurement accuracy. In this study, the cracks of a reservoir dam are taken as objects, and a new crack backbone refinement algorithm and width-measurement scheme are proposed. The algorithm simplifies the redundant data in the crack image and improves the efficiency of crack-shape estimation. Further, an effective definition of crack width is proposed that combines the macroscale and microscale characteristics of the backbone to obtain accurate and objective sample points for width description. Compared with classic methods, the average simplification rate of the crack backbone and the average error rate of direction determination are all improved. The results of a series of experiments validate the efficacy of the proposed method by showing that it can improve detection automation and has potential engineering application.