Track inspection is a critical process for both freight and transit rail safety. Currently, many rail transit agencies rely heavily on visual inspection, which is time and resource consuming, inefficient, and sometimes unreliable. Although some rail transit agencies do use ultrasound or track geometry car for track inspections, these devices can only identify certain types of defects and are usually expensive. Many small- or medium-sized transit agencies cannot afford to purchase them and only hire consultants to do inspections once or twice a year, even though they do have the needs to conduct the inspections more frequently. To address this issue, this research introduces a new computer vision framework specially designed for the automatic inspection of railroad tracks. The proposed framework is based on the Laser Crack Measurement System (LCMS), which consists of two high-performance 3D laser profilers that are able to measure complete transverse railroad profiles with 1mm resolution at high speeds. Compared to ultrasound and track geometry cars, the developed system is much more affordable and can be readily mounted on a high-rail vehicle that virtually every rail transit agency owns. Based on the 3D depth map generated by the LCMS, new methods have been developed for measuring rail gauge, detecting missing or broken fasteners, and identifying cracks in concrete ties. Preliminary results reveal that the newly proposed computer vision framework is a promising cost-effective and reliable alternative to the existing inspection methods.

View Paper