Predicting London's air pollution
with 'JamCams'
Air pollution is a major problem in London. It costs the city £16 billion and leads to 9,400 premature deaths per year. Current Mayor, Sadiq Khan, has been stating that “The shocking and illegal state of London's filthy air" has forced him to not only trigger alerts, but also to allocate £20 million to achieve the target of zero vehicle emissions by 2050. Many programs have been launched to address this, but air quality measurements to assess their impact are sparse and have significant gaps.
As part of the S2DS London 2017 program, HAL24K launched a data science project to demonstrate that traffic cameras could help in air quality forecasting. The team used deep learning and image recognition to identify vehicles in the traffic camera data, while machine learning techniques made predictions of air quality. We developed a pipeline for the processing of traffic videos that were then fed into a model to determine traffic intensity before deploying the system at scale in the cloud. A second model was generated for the prediction of air quality, after which a visualization was created of air quality and traffic in London. The project proved a breakthrough in increasing the understanding of the most polluting vehicles and the affected areas
HAL24K is now actively building upon the team's work to create an even more powerful solution that functions within our Dimension data science platform. As we incorporate more data and cameras into the model, we can see how the impact of traffic on air quality varies across London. These predictions can be used to better target interventions such as traffic flow restrictions and public transport improvements, lowering pollution overall and improving the health of London's inhabitants.