Real-time Vehicle Speed Detection Based on YOLOv11 and DeepSORT

The technology of real-time and effective vehicle speed detection is considered a key technology to improve traffic monitoring efficiency and traffic safety management grade. To address the limitations of traditional speed detection schemes—including reliance on dedicated hardware, poor environmental adaptability, and high construction and maintenance costs—this paper proposes a r10eal-time vehicle speed detection system based on YOLOv11 and the DeepSORT algorithm. The proposed system uses the YOLOv11 target detection algorithm as its primary model. DeepSORT multi-target tracking technology is integrated to enhance tracking performance. Speed measurement is implemented using a virtual detection line. This approach enables accurate vehicle detection, continuous tracking, and real-time speed measurement within video frames. Through the experiments, the result shows that the improved YOLOv11n model reaches mAP@0.5 of 0.982 and a recall rate of 0.956 in the test set, higher than the YOLOv8n and YOLOv5s models. The speed detection error can be restricted to 3 km/h, satisfying the real-time detection need. There is no need for road surface modification, and the detection system has a flexible layout, providing a dynamic basis for traffic law enforcement and traffic data support for road construction and traffic control optimization.