PERFORMANCE OPTIMIZATION OF ADAPTIVE TRAFFIC LIGHTS USING MACHINE VISION

Abstract

The paper analyzes road and pedestrian traffic at a controlled intersection using neural networks in the tasks of interpreting video stream coming from street surveillance cameras. A new approach for optimizing node throughput is given based on intelligent technology for interacting the infrastructure with I2P transport and road traffic. The factors that affect the reduction of the efficiency in the use of road infrastructure have been identified. On the basis of dynamic monitoring of road and pedestrian traffic, algorithms for the operation of traffic lights have been developed, taking into account the parameters of pedestrian traffic. The solution is based on the collection and processing of dynamic data of road and pedestrian traffic in real time for the adaptive traffic light training system. The “smart traffic light” is based on the principle of creating minimal impacts on pedestrian traffic and ensuring maximum traffic capacity when turning to the right. The study of the given approach is conducted at one of the busiest intersections in Chelyabinsk and can be used at other nodes of the city's road network (RN).