Modified Hough Transform-Based Registration Method for Heavy-Haul Railway Rail Profile With Intensive Environmental Noise

In heavy-haul railway inspections, accurate rail profile registration is essential for assessing wear and deformation of tracks. In this process, identifying the rail waist arc as key segments and determining the center of the arc as feature points are crucial steps for rough rail profile registration, often affected by noise from sensors, surface contaminants, and environmental factors. To address these challenges, a profile registration method is proposed, with the modified Hough transform (MHT) as a core component. The method first segments key profile sections using curvature thresholds to identify arc segments. For each point in these segments, MHT constructs circles with fixed radius and calculates their intersection points. A distance matrix based on Euclidean distances between intersection points is generated. By sorting and normalizing these distances, the method identifies local density variations and calculates density coefficients to find the circle center, without manual thresholding. Comparative experiments on simulated and field data show that the proposed method achieves high accuracy and robustness in noisy environments. The dual validation confirms its reliability, making this method effective for rail profile registration in heavy-haul railway inspections.