Researchers at South Ural State University (SUSU), a partner of TV BRICS, have patented a programme to detect traffic anomalies using neural network technology. It analyses CCTV footage, recognises different vehicles and tracks their speed and trajectory with an accuracy of up to 30 centimetres. This allows to form visual maps of traffic flow difficulties in real time, the press service of the university reports.

According to Olga Ivanova, associate professor of the SUSU Department of System Programming, the peculiarity of the programme is that it can detect even minor deviations in traffic, including a reduction in lane width. The system signals possible obstacles, such as traffic accidents or repairs. The visualisation of intersections is updated every two seconds, with a colour-coded indication of congestion – the more cars in an area, the redder the box.

It is noted that in the future, the neural network will not only be able to detect anomalies, but also classify them by type, predicting possible congestion and its impact on traffic in the next 10-20 minutes. This will help the relevant departments and services to promptly respond to road problems and avoid worsening the situation.

According to Ivanova, the development has an important advantage, it is easy to integrate into the infrastructure of any city, and it does not require the installation of expensive GPS sensors on each vehicle.

The project is part of an intelligent transport monitoring system that is already operating in several major Russian cities. The introduction of the new programme is expected to improve the transport situation in cities by increasing traffic speed and reducing the environmental load.

Text copied from https://tvbrics.com/en/news/russian-scientists-teach-ai-to-detect-traffic-anomalies-on-roads/