Integrating AI and edge computing for advanced safety at railroad grade crossings

The integration of artificial intelligence (AI) and machine learning (ML) into the Railroad High-Grade Crossing (RHGC) systems represents a significant milestone in enhancing both safety and operational efficiency. The fusion of various technologies enables the seamless integration of real-time identification, accurate forecasting, and prompt reaction to pivotal traffic situations. This research introduces a state-of-the-art architecture based on edge cloud technology. It integrates advanced computer vision algorithms for object detection and segmentation with a custom dataset for RHGC safety. In the proposed novel framework, the authors utilize a Weighted box- Fusion (WBF) ensemble approach, integrating diverse object detection algorithms, such as YOLOv8M (medium), YOLOv8L (Large), and YOLOv8X (extra-large), to enhance the detection of safety measures at RHGCs objects. Moreover, the authors incorporate a UNet segmentation model to identify trains approaching the RHGCs. The amalgamation of these methodologies leads to a fully automated, AI-driven safety mechanism for RHGC. The edge-cloud architecture is employed, with surveillance cameras linked directly to an edge server strategically positioned at grade crossings. This arrangement facilitates real-time data processing, ensuring efficient bandwidth usage and minimal latency by relaying only the necessary processed information to the cloud. The authors' ensemble model demonstrated an impressive precision rate of 97%, with the segmentation model achieving a higher rate of 98%. This system establishes a novel standard within the discipline, amalgamating artificial intelligence, edge computing, and cloud technology to significantly augment safety and efficiency at grade crossings.

• Record URL:
  • https://doi.org/10.1016/j.jrtpm.2024.100501
• Record URL:
  • http://www.sciencedirect.com/science/article/pii/S2210970624000714
• Availability:
  • Find a library where document is available. Order URL: http://worldcat.org/issn/22109706
• Supplemental Notes:
  • © 2024 Elsevier Ltd. All rights are reserved, including those for text and data mining, AI training, and similar technologies. Abstract reprinted with permission of Elsevier.
• Authors:
  • Amin, A L
  • Chimba, Deo
  • 0000-0002-2881-1417
  • Hasan, Kamrul
  • 0000-0002-1844-3569
• Publication Date: 2025-3

Language

• English

Media Info

• Media Type: Web
• Features: Figures; References; Tables;
• Pagination: 100501
• Serial:
  • Journal of Rail Transport Planning & Management
  • Volume: 33
  • Issue Number: 0
  • Publisher: Elsevier
  • ISSN: 2210-9706
  • Serial URL: http://www.sciencedirect.com/science/journal/22109706

Subject/Index Terms

• TRT Terms: Cloud computing; Computer vision; Machine learning; Object detection; Railroad grade crossings; Railroad safety; Traffic surveillance
• Geographic Terms: Nashville (Tennessee)
• Subject Areas: Data and Information Technology; Planning and Forecasting; Railroads; Safety and Human Factors;

Filing Info

• Accession Number: 01943923
• Record Type: Publication
• Files: TRIS
• Created Date: Jan 27 2025 3:39PM