Research on Real-Time Monitoring of Vehicle Queue Lengths at Highway Toll Stations Using YOLOv5 and Kalman Filter Tracking Algorithms

Dynamic monitoring of queue lengths of vehicles waiting for toll booths on highways is critical for maximizing traffic flow and increasing traffic performance. On the other hand, traditional methods mainly utilize fixed sensors that have various issues including high cost and low flexibility. To address this problem, this paper introduces a novel model based on the YOLOv5 object detection algorithm and Kalman filter tracking algorithm to achieve real-time monitoring of vehicle queue length. First of all, the novel model utilizes YOLOv5 to accurately detect vehicles and get each vehicle’s bounding box information. Then Kalman filter algorithm is used to predict and track the motion state of the vehicle, and the position and speed of each vehicle are estimated accurately. The model calculates queue length in real-time by continuously monitoring the position and speed of each vehicle. To improve the complexity and accuracy of the model, a multi-target tracking framework is introduced to avoid the occlusion and interaction between vehicles, improve the stability and accuracy of tracking. The experiments show that the proposed model has high accuracy in vehicle detection and target tracking and that the processing time is significantly reduced compared with the traditional methods, which has better performance.