Published 27 Jan 2022 by Mitt i Stockholm
Reporter: Ingrid Johansson. Photo: Sacharias Källdén
Platforms at 14 stations have AI cameras that recognize deviant behaviors in travelers – for example, if someone is drunk, or is about to jump onto the tracks.
Big brother sees you – and warns if you seem to end up on the track. Region Stockholm’s attempts with AI surveillance on metro platforms have resulted in two suicide attempts being prevented.
“These are people who jumped onto the track and where they managed to stop the trains in time”, says public transport regional councilor Tomas Eriksson (MP).
Sends alarms for deviant behavior
The surveillance cameras on the platforms have been equipped with AI programs from Irisity. These recognize how travelers behave and react when someone deviates, for example, staying on the platform for a prolonged time.
It can also have to do with intoxicated people who get too close to the edge of the track, or about track running. The cameras then alert the security center, who can act by announcements into a speaker or sending safety staff over to check out the situation.
“On 11 occasions so far, the alarms has been regarding suspected suicide attempts. Staff were called to the iste and were able to help the persons involved”, Tomas says.
Most deviations in Vällingby
Currently, the technology has been going on for almost three months. At that time, 276 alarms were received. 108 of them were people who came too close to the edge and 91 were people present on the tracks.
Of the 14 stations, it is in Vällingby that the cameras have had to work the hardest, with 47 alarms. Then comes Liljeholmen with 40. Unsurprisingly, it is during Saturday and Sunday, and between 10 pm and 1 am, that most deviations are reported.
“This saves lives. In addition, unauthorized track running is minimized, which results in fewer disturbances in traffic. The ambition is to install this at all stations”, says Tomas Eriksson.
Throughout 2021, 154 suspected suicidal people were taken into custody on the subway. The experiment with AI monitoring is ongoing.