Determining the Carbon Pollution in a City by Monitoring Traffic Flows Using Machine Vision

The article discusses a method for determining the level of carbon pollution in a city from automobile emissions in relation to the current degree of congestion of a section of the road network. Information about traffic jams is taken from online street surveillance cameras and processed by the yolov5 object detector, trained on a generated and labeled dataset, including the movement of cars on city roads. The detector model makes it possible to determine with a high degree of accuracy the coordinates of the bounding boxes for most cars on a section of the road, thereby recording the fact of their movement, as well as determining their average speed. This information is used to calculate the intensity and density of vehicle traffic over a certain time on a selected section of a four-lane road. Certain traffic flow indicators are used to calculate the amount of carbon dioxide emissions in relation to engine power and specific fuel consumption at the corresponding current speed and engine speed. The resulting values are then converted into carbon units and displayed in real time on a digital map of the city. The map of carbon unit emissions can subsequently include layers with emissions from industrial enterprises, agricultural facilities and other production facilities and be used to justify and make management decisions to reduce the load on the environment of the city and region.