In recent years,provinces such as Zhejiang and Fujian have successively introduced relevant local regulations pro-hibiting two wheeled vehicles(such as motorcycles,electric vehicles,etc.)from passing through highways.A modified YOLOv7-Tiny two wheeled vehicle intrusion detection algorithm is proposed to address the issue of real-time detection of two wheeled vehicle in-trusion by highway entrance workers.Firstly,motorcycle images were extracted from VOC2005 and images with entrance back-grounds were added to form a new dataset.Secondly,based on YOLOv7-Tiny,an ECA attention mechanism was introduced to make the model more focused on training motorcycle related target features.The ssFPN network was used to enhance small target feature information,and a WIoU loss function based on dynamic non monotonic mechanism was used to improve the accuracy of small ob-ject detection.Finally,use the Adam optimizer to improve the convergence speed and accuracy of the regression process.The im-proved algorithm improves mAP,Precision,Recall by 2.63,4.01 and 13.92 percentage point,respectively,and improves F1 by 0.10,indicating significant effectiveness of the method.
two wheeled vehicle intrusion detectionYOLOv7-TinyECA attention mechanismssFPNWIoU
NETL
NSTL
万方数据
