Structured-Light Vision Measurement Approach for Geometric Parameters of Wheelset Under Variable Re-Profiling Working Conditions

Re-Profiling is the technique that restores worn wheelsets to operational standards, with the selection of re-profiling parameters based on the measurement results of wheelset geometric parameters. Measurements in the variable re-profiling conditions exhibit the following difficulties: limited space for sensor installation; requirement of multiple repeated measurements before and after re-profiling; and complex diffuse reflection on the wheelset surface after re-profiling. Owing to the above reasons, the existing measurement system cannot be applied under the wheelset re-profiling conditions. This paper proposed a structured-light vision measurement approach and developed an efficient system for measuring the geometric parameters of wheelsets in variable working conditions during re-profiling. Initially, a stripe extraction method in adaptive ROI (region-of-interest) based on deep learning is proposed to achieve high-precision extraction of laser stripes. Then, two structured-light sensors are configured as a binocular system, and a 2-D image profile registration method based on binocular vision is proposed to accomplish the coordinate unification of the measured profile. The feasibility of the measurement is verified by analyzing the possible influencing factors as well as practical experiments, the results demonstrate that the maximum error of on-site measurements in before re-profiling stage and after re-profiling stage are 0.35 mm and 0.43 mm respectively, which meet the requirement of 0.50 mm.