Traffic Flow Detection Based on Improved YOLOv5 and DeepSORT

Intelligent Transportation Systems (ITS) are vital for managing traffic in high-flow areas like high-ways and city centers. Our approach to ITS involves real-time road monitoring using cameras and periodic video saving, followed by object detection and tracking with an improved YOLOv5 and DeepSORT algorithm. This combination provides high precision in traffic flow detection, essential for effective traffic management. The YOLOv5 algorithm is renowned for its speed and accuracy in object detection, while DeepSORT excels in tracking objects over time, even when they temporarily disappear from view. Our enhancements to YOLOv5 focus on its suitability for real-time traffic monitoring, ensuring high precision without excessive computational demands. Our experiments on the UA-DETRAC dataset, which offers a realistic and challenging environment, demonstrated a 70.1% mean Average Precision(mAP) at a 50% Intersection over Union(IoU) threshold. This achievement highlights the effectiveness of our algorithm in real-time object detection. The lightweight design of our algorithm is crucial, as it allows for deployment across various devices without compromising performance. This is particularly important for real-time applications, where quick data processing and analysis are necessary for timely insights. The high precision of our traffic flow detection algorithm is a significant advantage, providing valuable data for traffic management decisions. It can aid in adjusting traffic signals, implementing congestion pricing, or deploying emergency services in response to accidents. In summary, our algorithm represents a significant advancement in ITS, offering a high-precision, real-time, and light-weight solution for traffic flow detection. Its successful application on the UA-DETRAC dataset shows its potential for real-world deployment, making it a promising tool for enhancing traffic efficiency and safety in urban environments. As traffic demands grow, the integration of such intelligent systems will be key to shaping the future of transportation.