3D reconstruction of track slab sides based on a moving single-linelaser system and precise registration

High-precision 3D reconstruction is a core technology for visual inspection of rail track structures, playing a vital role in the timely detection of rail defects. Existing structured light 3D reconstruction methods struggle to adapt to field scanning tasks for track slabs, a large-scale infrastructure, especially the sides of the slabs, where many concealed defects endangering safe operations exist such as seam gaps, spalling, and cracks at wide-narrow joints. Addressing this challenge, a high-precision 3D reconstruction method for track slab sides based on single-line laser mobile scanning is proposed in this paper. For capturing sequential point clouds of track slab sides appropriately, we design a single-line laser dynamic scanning system comprising a 3D camera and a single-line structured light sensor first. Then, a globally optimized camera pose estimation algorithm based on spatiotemporal constraints, termed Pose-STC, is proposed. Specifically, temporal constraints enforce smoothness in translation and rotation between adjacent camera poses while spatial constraints incorporate the geometric priors of the track slab by ensuring spatial consistency of extracted structural feature points. Further, we propose a motion compensation strategy based on non-linear optimization for accurate registration frame by frame to achieve high-precision 3D reconstruction of the slab side. Comprehensive laboratory and field tests demonstrate that the proposed method delivers excellent performance, surpassing conventional methods in challenging field environments. Both coplanarity and collinearity measurements achieve sub-millimeter accuracy, meeting the precision inspection requirements for track slabs.