Assessing the Effectiveness of Vision Technologies for Railcar Inspection
- Language of the publication
- English
- Date
- 2024-12-02
- Type
- Report
- Author(s)
- Yan Liu
- Abdelhamid Mammeri
- Samy Metari
- Md Atiqur Rahman
- Alireza Roghani
- Michael Hendry
- Lianne Lefsrud
- Parth Rana
- Fereshteh Sattari
- Publisher
- National Research Council Canada
Alternative title
Évaluation de l’efficacité des technologies de vision pour l’inspection des wagons de chemin de fer
Abstract
Railroads worldwide are leveraging machine vision technologies to enhance railcar inspection quality and efficiency, ultimately improving railway safety. In collaboration with Canadian Pacific Kansas City (CPKC), the University of Alberta (U of A), TC’s Rail Safety and Security Directorate, and the National Research Council Canada (NRC), Transport Canada’s Innovation Centre (TC) launched the Automated Machine Vision Inspection Systems (AMVIS) project in 2021 to assess the capabilities of remotely monitored train inspection technologies. This project studied the reliability of the Train Inspection Portal System (TIPS) under various climatic conditions, the effectiveness of Portal Office Inspection (POI) in detecting safety defects, and the potential of AI algorithms to support inspectors. The results provide evidence that TIPS enables real-time, high-quality imaging without disrupting train operations, reduce idling time and improve defect detection for several defect types. Additionally, AI models such as YOLOv5 and Faster R-CNN demonstrate strong potential in automating defect identification, particularly for wheels and cap screws. Key recommendations include optimizing camera placement for enhanced imaging, refining POI software for improved defect detection, and integrating AI-driven solutions to elevate inspection accuracy and efficiency.
Description
This report investigates the reliability of the Train Inspection Portal System (TIPS) under different climatic conditions and assesses the effectiveness of Portal Office Inspection (POI) in detecting train defects. It also examines the impact of human factors on POI output and explores the potential of AI algorithms to assist office inspectors.
Subject
- Rail transport
Keywords
- Rail Transportation,
- Automated Machine Vision Inspection System (AMVIS),
- Inspection Portal,
- Machine Vision,
- Human Factors,
- Defect Simulation,
- Train Inspection,
- Train Car Defects,
- Artificial Intelligence
Rights
Pagination
1-89
Peer review
Internal Review
Open access level
Green
Identifiers
- Government document number
-
https://open-science.canada.ca/handle/123456789/3472
- 1TJC06XY54SB-1102879366-1463
Report
Citation(s)
Liu, Y., Mammeri, A., Metari, S., Rahman, M.A., Roghani, A., Hendry, M., Lefsrud, L., Rana, P., Sattari, F., (2024). Assessing the Effectiveness of Vision Technologies for Railcar Inspection, National Research Council Canada, University of Alberta.