The CCTV system in Moscow can now recognise faces. It uses an algorithm based on neural networks: city camera recordings are analysed in real time mode. Faces on the screen are scanned and can be checked against several data bases, for example, the police data base to identify a suspect. This analytical system can also help police recreate a suspect’s movements around the city. The system will search for related recordings from various CCTV cameras and identify the same face from several sightings. The Moscow network includes 160,000 CCTV cameras and 95 percent of residential buildings. By the end of the year, residents will be able to install CCTV cameras on their buildings themselves and then connect them to the unified video observation system.
“The introduction of video analysis is a powerful tool for increasing the effectiveness of the city’s monitoring systems as well as for private applications. Muscovites now have an additional level of security,” said Head of Moscow’s Department of Information Technology Artyom Yermolayev. “Of course these opportunities need to be introduced with caution. Our priority is to maintain a balance between privacy and security, and we comply with the strict policy for control and respect for an individual’s rights.”
A department spokesperson added that Moscow has one of the largest security systems in the world with such a comprehensive identification system.
Now about 16,000 users are connected to the city’s security system: the police, and federal and regional security agencies. Each jurisdiction has its own access level, which helps maintain information confidentiality. The police must request specific information, under current law, and federal agencies can only access CCTV cameras from the areas and routes they are responsible for. Every request in the system is recorded.
Face recognition works online, with identification taking only seconds. If the algorithm finds someone whose face has a match in a data base, it will notify the law enforcement agencies.
The department spokesperson also notes that face recognition has already increased the efficiency of investigations and suspect searches. During the pilot programme over 50 suspects were identified and arrested with the help of the analytical algorithm. Some had been hiding from police for years.
The public will be able to connect their CCTV cameras to the common city network. This option will be introduced by the end of this year. The video from these cameras will be uploaded to the general centre for data storage and processing (GCDSP), and the recordings can be used as evidence in court.
This year, over 3,500 cameras have been connected to the GCDSP. This includes cameras in entrance halls, in schools and kindergartens, at MCC stations, stadiums, public transport stops, bus terminals and in parks. Also, cameras will be installed in 25 underground pedestrian crossings by June 2018. The devices will appear in underground passages not connected to metro stations or supervised by Gormost (bridge and tunnel authority).