Automatic contour registration method in rail abrasion dynamic measuring

Abstract

The invention provides an automatic contour registration method in rail abrasion dynamic measuring. The method includes: using an RDP algorithm to automatic segment an obtained rail web and a rail bottom arc contour; performing nonlinear quadratic fitting based on radius constraint on the segmented arc contour to obtain circle center coordinates, and using the circle center coordinates as the reference point to perform primary registration on a rail measured contour and a standard contour; using the primary registration result of the reference point as the initial value of an improved ICP algorithm, and performing iterative computation to obtain the optimal rigid body transformation relation to complete the automatic contour registration in rail abrasion dynamic measuring. By the two-section automatic rail contour registration method based on the RDP algorithm and the improved ICP algorithm, high-precision automatic registration of the rail measured contour and the standard contour isachieved.

Description

technical field [0001] The invention relates to the technical field of rail transit infrastructure safety detection, in particular to an automatic registration method for rail profiles in dynamic measurement of rail wear. Background technique [0002] Rail is an important infrastructure supporting the operation of trains. The high-speed, heavy-load, and high-density operation of trains has made the problem of rail wear increasingly prominent. Excessive wear of rails will cause great harm to the stability, safety, ride comfort of trains and the service life of various parts of vehicles and tracks. Accurate measurement of rail wear is of great significance for formulating maintenance and repair plans and ensuring railway traffic safety. [0003] For a long time, for the measurement of rail wear, the railway department mainly uses special calipers to conduct regular sampling inspections manually. This detection method is inefficient and cannot realize dynamic measurement, and ...

AI Technical Summary

Problems solved by technology

At present, the Gauss-Newton nonlinear quadratic fitting method based on least squares is used to obtain the coordinates of the center of the rail waist arc as feature points, and the registration is performed through affine transformation, but the initial segmentation of the contour depends on the coordinates of the sensor used It is easy to introduce errors; also use the arc center of the rail waist as the reference point for registration, and determine the tangent point of the arc segmentation by setting the curvature threshold. This method can realize the rail waist, rail The automatic division of the tangent points of different arcs at the bottom, but the curvature fluctuation is large due to noise interference in the actual measurement, it is difficult to determine a stable segmentation threshold; the iterative closest point (Iterative Closet Point) algorithm is used to determine the distance between the measurement coordinate system and the design coordinate system Only use the same sensor to realize the reference measurement and wear measurement. However, in order to improve the efficiency of the algorithm, the range of iterative search is limited to a smaller window area, which makes the algorithm less robust and easy to fall into local optimum. untie

Images