Intelligent Video Analytics technology developed by university researchers promises to improve safety at rail and metro platforms and help ensure trains run on time.

Developed as part of a one-year research project, which received around £85k funding from the rail safety organisation RSSB, the proposed solution can use existing CCTV cameras located on platforms and on-board trains to provide real-time information to drivers and other rail staff when there are passengers in specific risk areas.

In the UK the number of passengers hurt or killed is low. However, passengers getting on and off trains at platforms is a time that carries some of the highest risks of accident.

The system uses novel (patent pending) intelligent video analytics computer vision technology developed at Lancaster University in collaboration with Digital Rail Ltd, a Lancaster-based rail safety technology SME, and builds upon the researchers’ previous successful work in computer vision and intelligent systems.

The technology is able to autonomously detect objects that are different from the background - such as people, luggage, bicycles, pushchair, mobility scooters, etc. - and estimates their position in relation to the yellow line, rails, and coach doors. It processes this information across multiple cameras along a platform and within trains to provide valuable information to staff and passengers in real time.

The system detects when passengers have moved between the yellow line marked on the platform and the train. The driver is alerted and shown at which door the potential problem is occurring. They can use this information to check relevant cameras to see if anyone is trapped.

In addition, it can inform drivers of approaching trains if someone has fallen from a platform and onto the tracks.

Professor Plamen Angelov, Chair of Intelligent Systems at Lancaster University, said: “This system can help the working experience of drivers, going some way to relieve their stress and pressure by assisting them in looking out for risks around the train and making accurate decisions.

“Gap falls, and ‘trap and drag’ scenarios can be detected more accurately and by using this system there is an improved chance that drivers and platform staff will notice these incidents before it’s too late.”

The system can also measure the likely busyness of specific train carriage doors before a train has arrived at a platform. It does this by using cameras within carriages to identify how many passengers are waiting to get off the train. The system also uses cameras on station platforms to see how many people are stood waiting for the train to arrive.

By combining data taken from cameras on platforms and within carriages the system can build a dynamic picture of which doors are likely to be busy, causing delays, and which doors are likely to offer smoother boarding for passengers.

This information can be used to inform waiting passengers, possibly through illuminated signs using a traffic light colour system, of the best places to stand, or avoid, in order to board the train more quickly and helping to reduce the time the trains need to wait at the platform.

Train dwell times have a significant effect on punctuality and therefore the network capacity.

“Our system will help staff to maintain on-time services by shortening passenger boarding dwell times, improving the passenger experience, diminish station bottlenecks and further improving safety by reducing crowds at stations,” said Professor Angelov.

The technology, which runs on standard computing equipment and can easily be miniaturized has the capability of detecting both moving and stationary objects. This unique feature is critical because it can separate and identify different kinds of objects, such as luggage, bicycles and passengers that could cause congestion around doors.

The system has been developed and tested using archive real-life CCTV footage from platforms and carriages on the UK rail network. The researchers now hope to further develop their software by testing it in real-time on real platforms and trains.

Researchers on the project include Professor Plamen Angelov and Dr Gruffydd Morris of Lancaster University’s School of Computing and Communications, and Dr Howard Parkinson of Digital Rail Ltd, which is co-located at Lancaster University’s InfoLab21.