A hardware-free framework that turns an existing network of city traffic CCTV cameras into virtual air-quality (NO2) sensors. Deep-learning detection and tracking extract per-type road-user flows from video frames; each camera's multi-channel traffic stream over an hour is compressed into a reparameterization-invariant path signature; and these signatures, together with static urban and environmental factors, feed a spatiotemporal graph deep model that predicts ground-level NO2 at locations without physical sensors. The approach also exposes lagged traffic-to-pollution relationships (via Granger-causality and spatial-regression analysis), enabling city-scale, real-time pollution monitoring and evidence for adaptive policy.
Air pollution in cities, especially NO\textsubscript{2}, is linked to numerous health problems, ranging from mortality to mental health challenges and attention deficits in children. While cities glo…