Object Detection on Railway
Context
SHARE ON
Object monitoring on the railways has traditionally been carried out by staff who monitored the data of video surveillance cameras installed in the infrastructure and by drivers of locomotives and passenger trains during the trips. This method of object monitoring is outdated and non-innovative, as the collection and processing of railway environment monitoring data are costly in terms of time and manpower. For this reason, systems based on computer vision technologies are emerging in the railway market, which can detect various objects of the railway environment (people, animals, objects, cars, etc.) and process the collected data in real-time from video surveillance cameras. Such systems, through the use of artificial intelligence capabilities, to learn and develop processes, open up greater opportunities for mobility companies to ensure safety and performance in their industries.
Challenges and Directions
Current Situation
The current total length of network lines of the Lithuanian Railways (LTG) is 1910.7 km. Video surveillance cameras are installed in all parts of the infrastructure, from the train driver’s cab and train stations to level crossings. However, in most cases, video cameras are monitored only for important reasons (received a complaint about a violation or due to accidents). Due to the lack of technology, data is not regularly recorded and stored in the system, which prevents in-depth analyzes and additional actions to improve the current situation (for example, if the system could detect movement of people on the tracks at a certain location, it would be possible to reduce the permitted train speed on a certain section. Or if the system would detect wild animals gathering near the railway and send a signal to the nearest station and other train drivers about the current dangerous situation on the tracks). For these reasons, LTG is looking for solutions that can improve railway infrastructure and make better use of existing infrastructure, which would allow the company to plan infrastructure development, monitor animal migration trends, increase safety, record passenger behavior, and observe violations in the railway area.
Project directions:
1. Object Detection at Level Crossings
Currently, the video cameras at level crossings record the general view of the level crossing from both sides, traffic light signals, the pedestrian crossings, and car license plates. Using the existing trunk communication lines, the signals of the video cameras are transmitted to the video recording device installed in the communication room of the nearest station.
2. Object Detection from the Driver’s Cab of a Locomotive / Passenger Train
Currently, most locomotives and passenger trains have external video surveillance cameras mounted on the side mirrors. Also, technical inspections of locomotives and passenger trains are now being fitted with video recorders, which operate on a similar principle to passenger cars.
3. Object Detection at Passenger and Cargo Stations
All Lithuanian passenger and cargo stations are currently equipped with video surveillance cameras. These data are monitored regularly (and, if necessary, video-recorded) by station security staff.
4. Object Detection at Passenger Station Platforms
Most of the platforms of Lithuanian passenger stations are currently monitored by video surveillance cameras. These data are monitored regularly (and, if necessary, video-recorded) by station security staff.
Solutions
Object detection at level crossings
- Recording traffic violations and scanning state car license plates;
- Calculation of cars passing the crossing;
- Monitoring of car congestion after opening the barrier;
- Recording of stopped vehicles or other foreign objects in the crossing area.
Object Detection from the Driver’s cab of a Locomotive / Passenger Train
- Monitoring of the railway environment (objects, activities, etc.);
- Tracking the migration of animals around the railway and their direction of movement;
- Unauthorized places where the illegal railway crossing is most often recorded;
- Determining the condition of a train running on an adjacent track.
Object Detection at Passenger and Cargo Stations
- Calculation of passenger flows;
- Identification of abandoned items and baggage;
- Determining the behavior and movement directions of people in stations;
- Identification of outsiders in the station area.
Object Detection at Passenger Stations Platforms
- Monitoring the behavior and actions of persons on the platform;
- Identification of abandoned items and baggage;
- Detection of persons who are in an unauthorized railway area;
- Calculation of the number of passengers on the platform.