Binocular Structured Light Vision for Segment Misalignment Inspection in Urban Rail Shield Tunnels

The shield tunnel, as a critical infrastructure component of rail transit systems, is prone to segment misalignment, a common defect that can lead to segment deformation, cracking, and water leakage, thus compromising the tunnel’s safety. High-density point cloud data are acquired using structured light binocular vision. Through intelligent recognition of point cloud circumferential seam features, this method automatically locates discontinuities and analyzes the direction of the circumferential seam main axis. The segment misalignment values are then automatically quantified using the zonal average elevation calculation method, with quadratic surface fitting applied to batch compute the misalignment data. To address challenges such as limited camera field of view and the impact of overlap on registration accuracy, FPFH, RANSAC, and multi-stage ICP algorithms are employed for point cloud registration and stitching. Experiments conducted on the Nansha-Zhuhai Intercity Railway shield tunnel demonstrate that this method can achieve a detection precision of ± 0.2mm. Compared to 3D laser scanning methods, it offers advantages in terms of lightweight design and lower cost, thus providing valuable technical support for the intelligent operation and maintenance of rail transit tunnels.