Design of a Traffic Lane Congestion Monitoring and Control System using YOLO Neural Network Approach

Abstract

Considering the most essential component of the traffic system, the traffic signal; this study presents a traffic lane congestion monitoring and control system with the assistance of image processing. The research uses YOLO (You Only Look Once) algorithm to detect the presence of different classes of vehicles, and then determine the response of the different classes of vehicles to a defined traffic condition. The algorithm achieves this by using a dedicated signal switching algorithm to have a considerable high accuracy even at varying resolutions. The trained model had an entire system accuracy of 86%. The algorithm also has a system sensitivity of 93% and precision value of 84%. The trained YOLO model had an accuracy of up to 99% for close up high-definition images. The result acquired help in regulating green signal time in traffic intersection by providing a more efficient and accurate signal with response to density.