Linear blue-light optical inspection for fatigue crack detection in steel bridges

Abstract

Fatigue cracking is a critical concern for the structural integrity of steel bridges, particularly in enclosed steel box girders characterized by dim illumination, reflective metallic surfaces, and complex background interference. Under such conditions, low crack-to-background contrast and unstable imaging quality severely limit the robustness of automated crack detection and feature extraction. To address this issue, this study develops a linear blue-light optical inspection approach for fatigue crack detection in steel bridges. Benefiting from the short-wavelength and directional illumination characteristics of blue light, the proposed method enhances scattering sensitivity to fine crack discontinuities, suppresses background reflection interference, and improves crack contrast against metallic surfaces. A systematic comparison of representative industrial illumination configurations demonstrates that linear blue illumination provides superior crack discriminability and imaging stability under complex surface conditions. Furthermore, to mitigate residual interference from metallic textures, corrosion regions, and uneven reflections, a tailored feature enhancement strategy is introduced to strengthen crack-aware feature representations and improve segmentation robustness. Experimental results show that the proposed approach achieves an mIoU of 75.63% and an F1-score of 78.51%, outperforming several representative methods. The results demonstrate that integrating illumination optimization with data-driven feature enhancement provides an effective framework for fatigue crack inspection in steel bridges under low-light environments.