Abstract
In this work, we propose an application of on-road surveillance cameras for traffic light optimization. The experiments consist of two parts one is data collection and another is traffic light optimization. In the data collection part, data were collected from the main road with one conjunction. Vehicles and Pedestrians were detected using the Yolov3 method. For lane detection, the Hough transform was used. After collecting the data traffic lights were optimized by some rules. Moreover, due to lack of data and sensitivity, the Simulation of Urban MObility (SUMO) simulator was also used for experiments. We achieved traffic light optimization by 15 seconds in real-time experiments while in the simulator, the time difference between the normal and optimized situation was 129 seconds to clear all the vehicles.