Real-Time Speed Estimation in Urban Traffic using YOLOv8

Accurate vehicle detection and speed estimation are vital for efficient traffic management and safety in urban areas. Traditional methods face challenges in cost, scalability, and handling dynamic traffic scenarios. This study employs the YOLOv8 model for real-time vehicle detection and a K-nearest neighbors (KNN)-based tracker to assign unique IDs to vehicles, enabling dynamic tracking. Speed is estimated in kilometers per hour by analyzing spatial displacement over time, utilizing predefined road markers. The proposed system effectively handles multiple vehicles simultaneously and provides reliable feedback on traffic flow. Experimental results demonstrate improvements in detection accuracy and speed estimation, achieving an average accuracy of 81.7% and a false detection rate of 2.1%. This system contributes significantly to Intelligent Transportation Systems (ITS), offering scalable solutions for automated traffic regulation and real-time monitoring. Future improvements aim to enhance performance in adverse conditions and complex traffic environments.