Field Depth Development of Weld Images in Orthotropic Steel Decks Using an RGB Camera

Quantitative detection of weld fatigue cracks in orthotropic steel decks (OSDs) typically relies on calibration objects or depth camera-assisted technologies. However, due to the size and power limitations of emerging mobile inspection equipment, as well as the spatial constraints within OSDs, existing methods or devices struggle to meet practical requirements. This article proposes a systematic method for field depth development of weld images using a portable RGB camera. First, based on prior knowledge of the structural parameters of OSDs and robotic arm kinematics, the spatial relationship between the structure and the robotic arm is determined. Second, leveraging optical projection theory and geometry, a 2-D vertical projection solution is derived. Third, the 2-D projection is extended to three dimensions (3-D) to establish a depth field for weld images, transforming linear data into a comprehensive surface representation. Finally, the experimental results demonstrate that the proposed method achieves a maximum relative error of 10.47%, with 98.95% of absolute errors in the weld region remaining below 20 mm. The research offers an innovative field depth development method, which holds significant reference value for all monocular visual depth-building tasks based on given scenarios.