Machine vision crack for better traffic management

The machine vision system of the Transport Board, which manages public transport and traffic in Tallinn, was developed in the fall of 2018 in cooperation with Sifr OÜ. The main goal of the created Krat is to find out what the traffic load of the city of Tallinn is, i.e. how many cars drive in and out of the city in one day, so that, based on this information, decisions related to city transport planning can be made based on the data (e.g. parking issues, road construction, etc.). An alternative solution would have been to install expensive sensors in the consoles of the intersections or in the road, but a better solution was found - to take advantage of the possibilities of a few hundred traffic cameras installed in the city to collect this information using camera images.


A crate-based solution


For testing, three cameras were selected from Tallinn intersections, which trained the Krat so that it would be possible to count vehicles passing by the camera. Training the crate for counting was an important process in the development of the solution, as there was no ready-made product corresponding to the conditions of the city of Tallinn. As a result, the Kratt was obtained, which counts buses, regular vehicles, trucks and motorcycles with the help of machine vision, and hopefully, in the next stage, also pedestrians and light-duty vehicles.



The obtained results were good enough - in a situation where there was no previous information, there was now nearly 70 percent accurate traffic flow information for the selected area. The accuracy of the results is affected by different weather conditions, smearing of the camera image and fog, but despite this, the result obtained from the three camera images is an important value for the city and an important milestone in the expansion of machine vision to hundreds of other city traffic cameras.



The company implementing the project, Sifr, is a subsidiary of a Swedish company that offers various developments of artificial intelligence solutions. The company's main customers are companies from the German pharmaceutical sector, the Baltic telecom, financial and insurance sectors.