Robust Rail Profile Processing With Rapid Registration: Wear Measurement for Heavy-Haul Railways in Harsh Environments

With the increasing axle load and traffic volume of heavy-haul railways, accurate and real-time rail wear measurement is crucial for operational safety. While traditional manual detection is inefficient, line-structured light-based onboard systems often suffer from noise, sparsity, and distortion caused by coal dust, fastener interference, and high reflectivity, alongside computational latency. Existing methods frequently lack robustness in correcting abnormal profiles and suffer from low processing efficiency. To address these issues, the Robust Real-Time Rail Wear Onboard Measurement System (RTRW-OMS) featuring a three-stage innovative algorithm is proposed. First, an adaptive DBSCAN integrated with PCA and a Short-line approach is employed for the robust elimination of environmental noise clusters. Second, a novel spatial continuity assessment based on 2D Fast Point Feature Histograms (FPFH2D) is developed to precisely identify and compensate for distorted profiles. Finally, an Improved Anderson-Accelerated ICP is utilized to achieve real-time wear calculation. Field tests validate that RTRW-OMS achieves a 99% denoising rate, a 78.3% anomaly identification rate, and a processing speed of 25ms per frame, significantly outperforming existing methods in robustness and efficiency.