YOLOv8-Based Framework for Vehicle Detection in Intelligent Traffic Monitoring Systems

Abstract

Traffic congestion and road accidents remain major challenges in modern transportation systems, demanding accuratereal-time traffic-monitoring solutions. Traditional surveillance methods and classical computer vision techniques are often limited by manual intervention, sensitivity to environmental variations, and poor scalability in complex traffic scenes. To address these limitations, this study proposes a YOLOv8-based framework for real-time vehicle detection in intelligent traffic monitoring systems. The proposed approach exploits YOLOv8 architecture to accurately detect vehicles of varying sizes under diverse traffic conditions, including occlusion, dense traffic, and illumination changes. A dataset comprising over 5,000 annotated traffic images representing urban and highway environments was used for training and evaluation. Model performance was assessed using standard object detection metrics, including Precision, Recall, F1-score, and mean Average Precision (mAP). Experimental results demonstrate strong detection accuracy, achieving an mAP@0.5 of 0.975 and a peak F1-score of 0.93, with stable convergence of training and validation losses. The results confirm the effectiveness of YOLOv8 as a lightweight and scalable solution for real-time intelligent traffic monitoring applications.