Sign up for our newsletter ⟩
News |

Helsinki looks at AI-enhanced camera system to count crowds

Test monitoring was carried out in downtown areas during August and September.

Kuvassa on Helsingin Senaatintorin jättiterassi heinäkuussa 2020.
A head count was made in Helsinki's Senate Square in August using video cameras and artificial intelligence. An algorithm calculated the number of people based on sample images. Image: Silja Viitala / Yle

The City of Helsinki says it is exploring the use of video surveillance tech that can accurately measure how many people are in public spaces.

Developed in cooperation with the information and security technology company Virian, the system tested is based on video analytics. It can determine how many people are on the move in the key outdoor spaces at different times of the day. Test monitoring was carried out in August at Helsinki's Senate Square and in September at Citizen's Square.

In practice, video cameras working in tandem with artificial intelligence measure the average number of people in an image area over a ten-minute period. After measurement, the information on the number of individuals in the space is stored in a database and a situation report is produced from the data.

Due to restrictions on crowd sizes aimed at preventing the spread of the coronavirus epidemic, the city wants the ability to keep track of congestion in public spaces.

According to Anssi Vuosalmi, the City of Helsinki's chief of preparedness, safe limit values were not exceeded at either pilot site during the test period. As there was a cap on gatherings of more than 500 people, monitoring numbers ranged from 1 to 500, mostly from a dozen to 200.

Vuosalmi said safety margins were maintained, explaining that the system used a wide-angle camera that takes a continuous video image of the area from which the calculation is made.

"For example, you can capture an area almost the size of Senate Square. The calculation can also be made from a smaller area when the image area is cropped," says Vuosalmi.

Crowds, not individuals

The City of Helsinki began looking at safety measures for mass gatherings in cooperation with rescue services and the police last year. The coronavirus epidemic accelerated piloting of the counting system.

"The coronavirus introduced a different perspective on the use of these kinds of calculations," Vuosalmi points out.

Primarily, as host to large gatherings, the city is looking for means to assess the size of crowds and anticipate safety requirements. For example, if emergency and rescue authorities set a certain maximum allowed number of participants, for example 10,000 people, this camera-based system could be used to monitor when that number starts to reach the upper limit. This would allow the rescue authority, or event organiser, to intervene before the area becomes congested.

As for privacy concerns, Anssi Vuosalmi says the system is secure because the AI handles the images as data blocks, not as photos of individuals.

"In addition, the camera image is not viewed by anyone other than machine intelligence that is merely processed by the algorithm. I only saw numerical values rounded off to the nearest tens," Vuosalmi says.

The City of Helsinki is now analysing the results of the tests and will make a decision in the future on implementing the public video headcount system.

The system that was piloted was developed and implemented by Viria's subsidiaries Viria Security Oy and Aureolis Oy.

Latest in: News


Our picks