Pedestrian Incident Detection in the Rail Right-of-Way using Artificial Intelligence

Details

• Creators:
  Cunningham, Chris ; Avr, Azhagan ; Searcy, Sarah ; Vaughan, Chris ; Coble, Daniel
• Corporate Creators:
  North Carolina State University. Institute for Transportation Research & Education
• Corporate Publisher:
  North Carolina Department of Transportation. Research and Development Unit ; North Carolina State University. Institute for Transportation Research & Education (ITRE)
• Corporate Contributors:
  North Carolina Department of Transportation. Research and Development Unit ; United States. Department of Transportation. Federal Highway Administration
• Subject/TRT Terms:
  • Artificial intelligence
  • Automatic data collection systems
  • Cameras
  • Pedestrian safety
  • Prototypes
  • Railroad safety
  • Real time information
  • Trespassers
  • Railroad Trespassing
• Publication/ Report Number:
  FHWA/NC/2019-50, NCDOT Project 2019-50
• Resource Type:
  Tech Report
• Geographical Coverage:
  North Carolina ; United States
• Edition:
  Final. January 2019 to June 2019
• Description:
  This research project builds on North Carolina Department of Transportation (NCDOT) RP 2015-18 (“Reduction in Railroad Right-of-Way Incidents”) and 2017-15 (“Rail Corridor Trespass Severity Assessment”). This effort focused on the development of a working prototype train-mounted camera system that will capture trespassing events in the nearby vicinity of moving or stopped trains. This dynamic system captures real-time trespassing data along any rail line, which will be used to better define trespassing issues. In the short term, the tools explored as part of this project will allow rail personnel to explain the extent of trespassing to municipal and law enforcement personnel, as well as the public. Prototype machine learning algorithms, sometimes referred to as “artificial intelligence”, were developed as a part of this project. The algorithms developed showed a lot of promise, even with a very limited library of thermal imagery in its database. Future research efforts should look to increase the image database to continue to increase the confidence in the algorithms ability to capture pedestrian events. Even with such a limited database, the team was able to capture a significant number of events on its test track.
• Format:
  PDF
• Download URL:
  https://rosap.ntl.bts.gov/view/dot/62732/dot_62732_DS1.pdf
• File Type:
  [PDF - 2.41 MB]
• Collection(s):
  US Transportation Collection
• Main Document Checksum:
  urn:sha256:7322ebb05f460e7158293942167e1d7c53d40467d05661d15fc64a54485cf6ed