DefinePK

Abstract

The pattern of change of conventional traffic lights does not consider the density of traffic in real time, thus hampering the efficient flow of traffic. Therefore, it is necessary to create and implement a more efficient control system that could maximize the flow of vehicular traffic. This paper proposes a method to enhance the regulation control of vehicles' density at an intersection by means of the dynamic estimation of the traffic light cycle using a deep convolutional based method. The proposed algorithm is oriented to estimate traffic lights' waiting time at simple intersections in real time. Once the video is processed, this approach can estimate a traffic light cycle based on an estimated traffic volume at a given time. All of the essential aspects of the methodology and materials used for the investigation are described. Algorithm improved the average queue length at intersections by 38% and improved the average waiting time by more than 60% compared with a traditional fixed-time cycle approach. Our proposal combines multiple ideas, image preprocessing, convolutional neural networks for object detection, and a traffic time estimation method based on Websters formulas. The proposed method, namely the dynamic estimation of the traffic signal cycle, showed a decrease in waiting times, the level of polluting emissions, and noise levels.