Feasibility study on the use of AI-based sensor technology for tunnel entrance monitoring
Funded by: Federal Ministry for Digital and Transport Affairs (BMDV) as part of the mFund innovation initiative
Project information
Funding period: September 1, 2023 - August 31, 2023
Project partners: iPattern Institute Hochschule Niederrhein, Masasana GmbH
Associated partners: Federal Police, DB Station & Service AG
Funding amount: 99,995 euros
Project vision
Deutsche Bahn owns more than 700 railroad tunnels. These must be protected against unauthorized access in order to maintain personal safety and operational safety. Due to the high number of tunnel entrances, constant by hand monitoring of the entrances is not economically feasible. Automated detection of intruders is made more difficult by the weather conditions and the changing light conditions at the tunnel entrances.
Existing systems
Existing monitoring systems based on conventional image cameras, thermal imaging cameras and LIDAR sensors have a high false alarm rate. As a result, the tunnel has to be closed and checked on site for intruders before it can be reopened to rail traffic. A precise detection system therefore not only contributes to safety, but also to the smooth running of rail traffic.
Project content
The project is being carried out jointly by the iPattern Institute of The Hochschule Niederrhein and Masana GmbH with the support of DB AG and the Federal Police. Masana GmbH is investigating automatic intrusion detection with color cameras, while the iPattern-Institute is investigating the use of Dynamic Vision Sensors.
A dynamic vision sensor differs from conventional cameras in that it does not capture complete images, but only changes in lighting conditions. These changes are passed on as events in the form "The pixel at position (x,y) has become brighter/darker". This type of data acquisition has several advantages in this application, including
- High temporal resolution
- Lower data throughput
- Restriction of the data to be examined to moving objects
- High effective contrast
- Improved data protection properties
The project investigated whether and how the use of these technologies can reduce the false detection rate without losing correct detections of intruders.
Information on the project results can be found in the final report and in the publication at ICPRAM 2024 entitled: "Intrusion Detection at Railway Tunnel Entrances Using Dynamic Vision Sensors".
