Cameras turn crowded train stations into useful data

July 30, 2021

Monitoring the crowds at train or subway stations with cameras isn’t new of course, but it has become extra relevant since the corona crisis. For example, large crowds can make it difficult to adhere to social distancing rules and creates greater risks for travelers. Because cameras can provide real-time insight into how busy a certain site is, you can then respond accordingly: by stopping escalators, for example, closing specific platforms, or, for example, asking travelers to leave half an hour earlier.

Mapping the crowds can be done by combining check-in and check-out data with image recognition, says Thales colleague Mathijs Voorend, transportation and ticketing expert. With current cameras, it is even possible to measure whether a distance of one and a half meters is being kept. We have actually been converting crowds into data for years: together with the Hong Kong company MTR, we developed a system a few years ago to use check-in and check-out data to determine in real time how many travelers are on a specific metro platform. “You know how many people fit in a metro, what the timetable is, whether a vehicle is delayed, and how long it takes to walk to the platform after checking in.” This way you can calculate how many people are waiting, how long their waiting time is and how many people have to stay longer because the first metro that comes along is full.

Traveler behaviour mapped

MTR learned more about the behavior of their travelers as a result. For example, the carrier saw that some travelers first traveled back two stops – in the wrong direction – to a quieter station. Voorend: “Apparently, that extra travel time is worth it for some to have a seat the next half hour.” Useful information for the carrier to better guide its travelers.

Of course, a Hong Kong metro station cannot simply be compared with Amsterdam Central, where train passengers can choose from numerous platforms from the same check-in gate. That is why Thales also developed a system with which cameras continuously measure how many people are on a certain part of a platform or in a station hall. For example, action can be taken when it is busy at the escalators, but not at the end of the platform.

No facial recognition

Incidentally, it concerns image recognition and not facial recognition, emphasizes Voorend. “It is purely an algorithm that recognizes and distinguishes a person from a trash can, dog or suitcase. It measures how many people there are per square meter and how close they are to each other.” In this way it can also be determined whether a distance of one and a half meters is kept. In addition, the cameras can recognize whether luggage has been left behind, whether there is aggression, and whether someone is wearing a face mask.

The two systems complement each other and together provide a good insight into passenger patterns. Check-ins do not say everything about the many platforms where travelers go, while cameras do not cover every corner. “Combined, you can easily determine where it is busy and at what times. You can obtain valuable information that data. It is up to carriers what they do with it.”

Secure sharing platform

During large crowds or events, there is often a need to share data with other parties such as the police. Or, as a transporter, with the help of cameras from the municipality around the stations and bicycle parking, they can gain insight into travelers heading to the trains. Thales developed the Secured Sharing platform for this, with which all this data can be safely brought together and then segmented.

“The NS, for example, supplies the check-in and check-out data, the municipality provides the traffic loop data around the station and ProRail measures how heavy the train is. If you combine that, you get an idea of how busy it is and will be," explains Voorend. According to him, the big advantage here is that the owner of the data is always in control of to whom, what kind of data, and when data is to be shared. This means conductors at Utrecht Central cannot view unnecessary information from Amsterdam Central, and that the police only gain access if something truly goes wrong.

This article previously appeared on in Dutch.