AI-enabled traffic cameras turn smog sensors, detect pollution-spewing vehicles in real time

Traffic cameras can now be used to detect city smog in real time, not only instantly identifying vehicles causing pollution but also replacing physical monitoring and cumbersome data analysis by humans on ground.

Scientists at the Council for Scientific and Industrial Research (CSIR) have developed an AI-Integrated Line Source Emission Inventory (AI-LSEI) dashboard that plugs into the existing CCTV cameras already watching our streets.

Instead of just looking for traffic violations, this AI tool identifies every vehicle that passes by, classifies it according to its type and calculates the level of pollution it is emitting, according to information shared by CSIR’s Science Communication and Dissemination Directorate (SCDD) on April 17.

Researchers at CSIR’s National Environmental Engineering Research Institute (NEERI) carried out several field trials, including one during a cricket match at the Vidarbha Cricket Association Stadium. During this, the dashboard recorded a massive spike. As fans poured in, carbon monoxide levels surged, accounting for over 57 per cent of the total emissions during the peak hours.

At another trial, the system was installed at NEERI’s own gate. From midnight to 4 pm on a typical Tuesday, the system recorded exactly 53 kg of carbon monoxide and 3.3 kg of particulate matter passing through that single point.

“Traditionally, creating a ‘pollution inventory’, a term for a detailed map of where emissions come from — was a slow, manual process. Traditional studies depend on manual counts and surveys, which can take up to one year to generate results,” according to Dr Sudip Karmakar from SCDD

“The AI-LSEI dashboard changes the game by working in real-time. It uses the live video feed to count traffic and then applies ‘emission factors’, scientifically established values that estimate how much chemical waste a specific type of engine produces per kilometer,” he said in a post on X.

The AI doesn’t just see a vehicle. It recognises categories ranging from two-wheelers to heavy commercial trucks and tractors. It even gets city specific data from the national Parivahan portal to understand the engine technology like BS-IV or BS-VI standards and fuel types common in the area, according to SCDD.

Within seconds, the dashboard calculates the ‘big four’ of urban air pollution – particulate matter (PM), which is tiny soot particles that can enter our lungs, nitrogen oxides (NOx), that are gases contributing to smog and respiratory issues, carbon monoxide (CO), a colourless, odourless gas that reduces oxygen delivery in the body, and hydrocarbons (HC), which are compounds that contribute to ground-level ozone.

The results are then splashed onto a GIS-based map, highlighting “hotspots” in glowing colours. If a specific corridor is choking under heavy truck traffic, the map turns red, giving authorities an immediate alert.

This tool isn’t just for scientists, but also for the people who run our cities. Smart City offices, traffic police and urban planners can use this data to decide where to divert traffic, where to synchronise signals to reduce idling or where to prioritise electric public transport routes.

Experts say that AI-LSEI is a viable pollution-monitoring system given the growing network of traffic cameras across the country. According to estimates, about 2.5 million AI-enabled cameras have already been installed in major cities and highways to check traffic violations and approximately 25 crore e-challans have been issued since 2019.