基于改进YOLOv5的电工着装检测方法研究

Research on Electrician Dressing Inspection Method Based on Improved YOLOv5

文章针对水电厂中工作人员着装不规范问题,提出一种基于改进YOLOv5的着装检测方法,即采用目标检测技术对工作人员是否佩戴安全帽以及工作服穿着规范与否进行自动检测。针对安全帽这类小目标检测问题,在YOLOv5网络模型的基础上嵌入轻量级ECAnet注意力机制模块,以减少无用信息通道的计算量,在保证YOLOv5检测速度优势的同时提高了小目标特征提取能力。结果表明,改进后模型的准确率、召回率、mAP@0.5分别提升了4.3%、2.1%、1.4%。

This paper proposes a dressing detection method based on improved YOLOv5 to address the issue of nonstandard dressing among working personnel in hydroelectric power plants.This method uses object detection technology to automatically detect whether working personnel are wearing safety helmets and their work clothes are wearing properly.For small object detection such as helmets,a lightweight ECAnet attention mechanism module is embedded on the basis of the YOLOv5 network model to reduce the computational complexity of useless information channels,while ensuring the advantage of YOLOv5 detection speed,the ability to extract small object features is improved.The results show that the accuracy,recall,and mAP@0.5 of the improved module increased by 4.3%,2.1%,and 1.4%respectively.