Neural network enters the city: how AI helps Moscow to monitor the environment

About the use of AI in environmental control of Moscow said the Deputy head of the Department of environmental management and environmental protection of Moscow Mikhail Ushakov.

Digital technologies accelerate the work of specialists

Under artificial intelligence in environmental control is understood several technologies. Video analytics recognizes objects and events on cameras, machine learning algorithms look for patterns in data sets, and generative models help specialists make the best decisions.

The system relies on data from cameras, sensors, camera traps and other means of objective control. Automation helps separate meaningful signals from digital noise.

According to Mikhail Ushakov, Moscow controls the state of air, water, green spaces and soil throughout the city. Objective control systems provide coverage unattainable only by human resources. Automated stations and autonomous sensors solve two tasks at once: they record violations in real time and serve as a source of statistical data. This allows you to model environmental and natural processes in the city.

So, 3 mobile environmental laboratories are involved for point control in the zones of influence of industrial enterprises and in problem areas. They operate 24/7, automatically analyzing the content of more than 10 priority pollutants in real time, and for an in-depth study of air quality, samples are taken for subsequent laboratory analysis of more than 120 indicators.

This integrated approach – a combination of stationary posts, mobile labs and predictive analytics based on AI – enables enterprises to quickly adjust technological processes and prevent disruptions before they occur. The introduction of such systems is an important step towards creating clean production and reducing the environmental burden on the city.

“Any artificial intelligence is built on data, so special attention should be paid to its quality, consistency and security. We are increasingly using technical complexes, sensors and cameras, and then process the received information using AI. This speeds up the work ten times, allows you to quickly identify possible violations and take measures, said Mikhail Ushakov.

Use of camera traps to prevent violations

Photo traps and sensors are installed where there are no city video surveillance cameras or it is difficult to conduct pedestrian surveys. The algorithm marks the frames where the dumping of waste or changes in the condition of the site are visible.

Information is checked by specialists and, upon confirmation of a violation, the necessary measures are determined. Artificial intelligence performs primary processing - around the clock analyzes thousands of images of the same type. At the same time, the final decision and responsibility for the conclusions remain with the specialists.

The data allow you to mark the bulk on the map and pre-select areas that require increased attention.

AI is also used in urban planning. Generative design allows you to compare improvement options taking into account the parameters of the territory, assessing their impact on the level of dust, noise, as well as temperature reduction in the heat islands.