基于改进YOLOv11算法的车辆类型及交通事故检测预警方法OA

To address the problem of false detections and missed detections in traditional detection algorithms due to insufficient detection accuracy for small targets and inadequate feature fusion in complex traffic environments,and to meet the transportation safety requirements of accurate identification of vehicles such as vans,trucks,and oil tank trucks as well as real-time accident early warning in such scenarios,based on YOLOv11,the detection and early warning technology for vehicles and accidents in complex traffic scenarios is studied,and an improved YOLOv11-based detection and early warning method is proposed.In the feature extraction stage,a SimAM attention module with slicing operation is designed,which slices the feature map into multiple subgraphs and applies SimAM attention to each subgraph respectively to enhance the local detail features of various types of vehicles;in the feature fusion stage,a content-guided attention(CGA)hybrid fusion scheme is constructed,integrating the advantages of the Feature Pyramid Network(FPN)and Path Aggregation Network(PAN),dynamically assigning weights according to the feature differences between vehicles and the equipment background to achieve efficient fusion of cross-scale features.The self-built dataset used in the experiment is based on the public traffic dataset from Roboflow,containing 6 832 images,covering various traffic hub scenarios,with target categories including van,accident,bus,car,oil tank car,and truck.The results show that the mean average precision(mAP@0.5)of the improved algorithm reaches 91.3%,and the detection precision and recall rate are increased by 2.1%and 1.9%,respectively,compared with the original YOLOv11.It can be effectively applied to numerous traffic scenarios,providing technical support for the safety management and control of traffic operations.