A Novel Probability Iterative Closest Point With Normal Vector Algorithm For Robust Rail Profile Registration

Regular inspection of rail wear is very important to ensure the safety of railway transportation. Rail profile registration is a key step in rail wear measurement. However, the accuracy of rail profile registration cannot be guaranteed due to soil and corrosion on the rail waist. To solve this problem, a new rail profile registration method based on probability iterative closest point algorithm with normal vector direction (N-PICP) is proposed in this paper. First, rail profile data is obtained using a line structured light measurement system. Then, the rail profile data is divided into rail head and rail waist according to the distance between adjacent points. The data of the rail waist is used to register with the corresponding part of the standard profile data by N-PICP. Finally, the rail head data is also converted according to the calculated transformation matrix. The experimental results demonstrate that the proposed method can realize the registration of the rail profile with a high precision and strong robustness.