基于改进的YOLOv4安全帽佩戴检测研究  

Research on detection of safety helmet wearing based on improved YOLOv4

作　　者：郭师虹[1] 井锦瑞 张潇丹 秦晓晖 GUO Shihong;JING Jinrui;ZHANG Xiaodan;QIN Xiaohui(School of Civil Engineering,Xi’an University of Architecture and Technology,Xi’an Shaanxi 710055,China)

机构地区：[1]西安建筑科技大学土木工程学院,陕西西安710055

出　　处：《中国安全生产科学技术》2021年第12期135-141,共7页Journal of Safety Science and Technology

摘　　要：为了加强建筑工人佩戴安全帽情况的检测,防止安全事故的发生,提出1种改进的轻量级YOLOv4安全帽佩戴检测算法,用于运行在移动设备端,降低现场部署的条件;制作1个8000幅图像的数据集,用于训练和评估安全帽检测算法;为了评估改进的YOLOv4的性能,从5个不同建筑工地采集到600张施工人员图像和60条施工视频作为验证集;根据建筑工地不同的视觉条件对图像进行分类,用于验证本文算法在不同外界环境下的性能。结果表明:改进后的模型检测速度是YOLOv4的3.4倍,可用于实时检测施工人员在不同施工现场条件下是否佩戴安全帽的情况,有利于提高安全检查和监督水平。In order to strengthen the detection of construction workers wearing safety helmets and prevent the safety accidents,an improved lightweight YOLOv4 algorithm for the detection of safety helmet wearing was proposed to run on the mobile devices and reduce the conditions for field deployment.A data set of 8000 images was made for training and evaluating the algorithm of safety helmet detection.In order to evaluate the performance of the improved YOLOv4,600 images of construction personnel and 60 construction videos from 5 different construction sites were collected as a verification set.The images were classified according to different visual conditions of the construction sites to verify the performance of the algorithm in different external environments.The results showed that the detection speed of the improved model was 3.4 times of YOLOv4,and it could be used to detect in real time whether the construction workers wear safety helmets under different construction site conditions,which is beneficial to improve the level of safety inspection and supervision.

关 键 词：施工管理 安全帽检测 轻量化神经网络 YOLOv4 实时检测 

分 类 号：X947[环境科学与工程—安全科学]