| 基于改进YOLOv5s算法的电动车头盔佩戴检测 |
 点此下载全文 |
| 引用本文：陆万浩,黄孙港，饶兴昌,孙曾阳.基于改进YOLOv5s算法的电动车头盔佩戴检测[J].计算技术与自动化,2026,(1):145-150 |
| 摘要点击次数: 190 |
| 全文下载次数: 213 |
|
|
| 中文摘要:为高效检测出路上电动车骑行者不按规定佩戴头盔的情况，提出了一种基于YOLOv5s模型的电动车头盔佩戴检测的改进算法。首先，该算法以YOLOv5s模型为基础，以C3Ghost替代原有的C3模块使得改进后的模型更加轻量化以减少参数量；其次通过嵌入Biformer注意力机制，并且将颈部网络中的Conv模块用GSconv代替；最后引入改进后的损失函数EIOU。通过消融实验对比后发现，改进后的模型算法平均精度均值(mAP)提高到了95.2%，精确率(Precision)提高到了95.6%。与YOLOv5s相比提升了16.9%和17%，同时浮点运算量下降25.5%左右。实验结果说明，改进后的YOLOV5s算法可以有效提升电动车骑行者头盔佩戴情况的检测性能。 |
| 中文关键词:YOLOv5 电动车头盔检测 轻量化 Biformer |
| |
| Research on Helmet Wearing Detection of Electric Vehicle with Improved YOLOV5s Algorithm |
|
|
| Abstract:In order to efficiently detect the situation of electric vehicle riders not wearing helmets according to regulations, an improved algorithm based on YOLOv5s model was proposed for helmet wearing detection of electric vehicles. Firstly,The algorithm is based on YOLOv5s model and replaces the original C3 module with C3Ghost to make the improved model more lightweight and reduce the number of parameters. Secondly, by embedding Biformer attention mechanism and replacing Conv module in neck network with GSconv.Lastly, the improved loss function EIOU is introduced. Through the comparison of ablation experiments, it is found that the average accuracy (mAP) and Precision of the improved model algorithm are increased to 95.2% and 95.6% respectively. Compared with YOLOv5s, it has improved by 16.9% and 17%, while floating point arithmetic has decreased by about 25.5%. The experimental results show that the improved YOLOV5s algorithm can effectively improve the detection performance of helmet wearing of electric vehicle riders. |
| keywords:YOLOv5 electric vehicle helmet detection lightweight Biformer |
| 查看全文 查看/发表评论 下载pdf阅读器 |
|
|
|
位访问者,今日访问
