Rail profile measurement based on structured light stereo vision

Traditional contact rail profile measurement methods suffer from low efficiency and lack continuous 3D data. Moreover, existing non-contact vision systems have limited ability to represent longitudinal features. To address these problems, this paper develops an innovative handheld high-precision 3D rail profile measurement system. The system combines multi-line structured light with binocular vision technology. By improving the Steger algorithm, it achieves sub-pixel accurate extraction of laser stripe centerlines. In the stereo matching stage, a novel combination of laser plane constraints and binocular left-right consistency constraints is applied. This effectively overcomes the negative impact of weak texture features on the rail surface for matching accuracy. The PnP algorithm is used for camera pose tracking to enable reconstruction of a complete 3D rail model. Experiments on a 60 kg/m standard rail validate the system. Results show an average 3D reconstruction error of 0.1 mm and measurement errors of key railhead dimensions below 0.04 mm. The study provides reliable and precise measurement technology to support intelligent operation and maintenance of railway infrastructure for enhanced performance.