We combined edge computing with state-of-the-art computer vision to estimate CO2 emissions of a traffic flow based on license plates without sacrificing personal information security. The proof-of-concept enables accurate estimation of emissions at road level while considering, e.g. shifts towards electric vehicles.
Cities have a natural incentive to reduce their carbon dioxide emissions caused by traffic, and monitoring the outcomes of policies requires measuring those emissions precisely. As vehicles have very different emission profiles, simply counting them on some roads is not enough. The license plate registry contains, in most cases, the CO2 emissions of the related vehicles but saving or sending license plate data anywhere would require treating it as personal data.
A computer vision application was created using the latest technologies to detect license plates from live footage. By combining the detected plates to emission data from the license plate registry, we could save and upload to the cloud only fully anonymous emission data without leaving traces of which vehicles were seen.