This system monitors traffic in urban context in real time and evaluates the vehicle flows. The video captured by a webcam is processed and the system is able to create dynamically a model of the roads (areas of the video frame where vehicle traffic is located) and calculate the vehicle traffic density for each flow.
Urban traffic monitoring is becoming increasingly important in order to optimize traffic, reduce travel times and reduce the environmental impact of traffic. Traditional techniques to count vehicles consist of installing mechanical or electrical sensors into the road pavement. These methods can be expensive and require physical changes (break pave, install sensors and repave the road). An alternative technique of traffic flow detection is based on video processing using computer vision algorithms. This technique is cheaper, more flexible and scalable than traditional ones.
The traffic flow detection algorithm has been developing from CSP in collaboration with the Formal Methods group of Department of Computer and Control Engineering ( DAUIN ) – Politecnico of Turin.
Different areas of the city with different road types (straight road, cross, roundabout) are used to test the system and a modified video streaming is created to show how the algorithm works:
A white line shows the backbone of the traffic area detected by the algorithm and each moving vehicle is displayed as a green spot. The algorithm creates a model of the roads through a static analysis, considering the areas of the video frame where there is an high number of moving object. For this reason in the roundabout of Collegno is monitoring only the left lane in the input road on bottom left.
All data collected by the system are sent to CSP IoT platform.
The algorithm is under evaluation to measure the accuracy of the data generated and to compare the results with other legacy algorithm.