Design and Implementation of Road Vehicle Counting Model Based on YOLOv8

With the gradual increase in the number of urban vehicles, traffic management is faced with some research and challenges. Therefore, a road vehicle counting model based on YOLOv8 is proposed to realize efficient and accurate vehicle detection and counting. Through calculations, the system can determine whether the road is crowded and then send signals to the traffic system to control the timing of traffic lights. This will reduce road congestion and traffic accidents. Specifically, we collected data from different cities and trained YOLOv8 models to track target vehicles using Kalman filtering tracking algorithms. To avoid counting inaccuracies due to irrelevant regions, we add irrelevant region masking. The results of the experiment were very promising, with the YOLOv8 model averaging relatively stable scores in accuracy, precision, recall, and f1 scores.