Key points are not available for this paper at this time.
The acceleration of urbanization and the consequent increase in population have exacerbated urban road traffic issues, such as congestion, frequent accidents, and vehicle violations, posing significant challenges to urban development. Traditional manual traffic management methods are proving inadequate in meeting the demands of rapidly evolving urban environments, necessitating an enhancement in the intelligence level of urban road traffic management systems. Recent advancements in computer vision and deep learning technologies have highlighted the potential of image processing and machine learning-based traffic management systems. In this context, the application of object detection and tracking technologies, particularly the YOLOv5 and Deep learning-based Simple Online and Realtime Tracking (DeepSORT) algorithms, has emerged as a pivotal approach for the intelligent management of urban traffic. This study employs these advanced object detection and tracking technologies to identify, classify, track, and measure vehicles on the road through video analysis, thereby providing robust support for urban traffic management decisions and planning. Utilizing digital twin technology, a virtual replica of traffic flow is constructed from camera data, serving as the dataset for training different YOLOv5 algorithm variants (YOLOv5s, YOLOv5m, and YOLOv5l). Upon comparison of training outcomes, the YOLOv5s model is selected for vehicle detection and recognition in video feeds. Subsequently, the DeepSORT algorithm is applied for vehicle tracking and matching, facilitating the calculation of vehicles' average speed based on tracking data and the temporal interval between adjacent frames. Results, stored in Comma-Separated Value (CSV) format for future analysis, indicate that the system is capable of accurately identifying, tracking, and computing the average speed of vehicles across various traffic scenarios, thereby significantly supporting urban traffic management and advancing the intelligent development of urban road traffic. This approach underscores the critical role of integrating cutting-edge object detection and tracking technologies with digital twin models in enhancing urban traffic management systems.
Kan et al. (Mon,) studied this question.