Moscow cameras are ceasing to be just "eyes" - now they also analyze what is happening. Artificial intelligence is increasingly integrated into the violation recording system, and the role of humans is gradually reduced to configuration, control and maintenance.
The scale is already impressive. There are about 3.7 thousand cameras operating in the capital, which record more than 60 types of violations. For 2025 - almost 37.7 million fines. And more and more of them are formed with the participation of neural networks.
Algorithms have learned to identify what inspectors used to record: unfastened seat belt, talking on the phone while driving, lack of a helmet. But the main thing is the fight against attempts to deceive. Even if the number is hidden or partially changed, the system is able to identify the car by external features - silhouette, color and details.
At the same time, there were some controversial situations. One of the most discussed cases is the fines received by the owners of cars transported by tow trucks. The cameras mistakenly read their numbers as violators. After that, a new rule was added to the system: if there is a tow truck in the frame, such cases are sent for additional verification.
AI not only records violations, but also helps to clarify them: if the number is not recognized, the system offers the inspector several possible matches of cars. The final decision still remains with the person.
Читайте ещё материалы по теме:
- Камеры научились «видеть» сквозь дым и пыль: новая разработка российских учёных
- Система «Безопасный город» поумнела: NtechLab установила 1600 камер с распознаванием лиц в Пензенской области
- Скачал «антирадар» — потерял деньги: появился новый вирус в Telegram