A Fast Rail Profile Point Cloud Matching Algorithm Based on Overlay Weight ICP

Abstract

The factors that affect rail's performance and quality mainly include fractures, abrasion, scratches, wear, etc. Among them, wear directly changes the profile of the rails. The first step in repairing the rail profile is to accurately measure the rail profile data, and then to calculate the wear of the rail by applying rail matching algorithm. In practical applications, the rail profile is variable and the measurement accuracy requirement is as high as ±0.2 mm. The matching algorithm needs to be able to adapt to various abnormal rail profiles without reducing the matching accuracy due to abnormal rail profiles. Therefore, designing with rail profile matching algorithm is highly difficult. The classic iterative closest point (ICP) algorithm is able to realize quick point cloud matching, but it has the shortages such as high requirements for initial input data, being prone to falling into local optimal solutions, and being susceptible to interference from noisy points, resulting in big matching errors. For the purpose of this article, an improved two-step rail matching algorithm was proposed. Firstly, the geometric features of the rail profile are used for fast rough matching, providing a high-quality initial input for the ICP algorithm. Then, the improved overlay weight ICP algorithm is used to quickly and accurately match the measured rail profile point cloud data with the standard rail profile point cloud data to calculate the rail wear data. The matching test results of various degraded rail profiles show that, for the wear at the top of the rail, the wear value calculated by the ICP algorithm with weights proposed in this paper is 0.763 mm less than the average error of the classical ICP algorithm. It has a good matching effect and can be used for the calculation of rail wear.

Keywords

• rail profile
• rail wear
• overlay weight
• iterative closest point (icp) algorithm