Traffic load - Machine vision for collecting information on traffic load

The Tallinn transport department, responsible for organising public transport and traffic in Tallinn, started developing a machine vision application in autumn 2018 in collaboration with Sifr OÜ. The main objective of this application is to identify the traffic load within Tallinn, i.e. how many cars drive in and out of the city in a day in order to make city transport planning related decisions based on such data (e.g. parking issues, road construction, etc.). Three cameras were selected on intersections in Tallinn to train the application to count the vehicles passing by the cameras. The training of the application was a crucial process in developing the solution since there was no ready-made solution available that suited the circumstances of Tallinn. The result is an application that uses machine vision to count buses, cars, lorries and motorcycles and in the future hopefully also pedestrians, cyclists and like.The result was information with a near 70% accuracy on the traffic load in a certain area when previously there no such information available at all. The accuracy of the results is affected by weather conditions, dirt on the camera lens, etc. Regardless, the results are a valuable milestone for the city for expanding the use of machine vision to other hundreds of city traffic cameras.