Research on highway hard shoulder running method based on YOLOv10s+SE attention mechanism

To alleviate highway traffic congestion during peak periods, this paper proposes a decision-making method for hard shoulder running based on the YOLOv10s object detection model enhanced with the SE attention mechanism. The proposed YOLOv10s+SE model is used to detect vehicles in real-time from surveillance videos, providing data on vehicle count, speed, and distribution. A traffic condition evaluation algorithm is then developed by integrating traffic flow, speed, and occupancy rate. Furthermore, a congestion probability prediction model and threshold-based decision mechanism are introduced to assess future congestion trends and determine whether the hard shoulder should be opened. Case study results show that the proposed method can accurately predict congestion and guide hard shoulder running decisions effectively, offering technical support for the intelligent transportation system.