主管：中华人民共和国应急管理部
主办：应急管理部天津消防研究所
ISSN 1009-0029  CN 12-1311/TU
消防科学与技术

消防科学与技术 ›› 2025, Vol. 44 ›› Issue (4): 541-547.

• • 上一篇    

深度学习下动火作业人员安全装备检测方法

赵鹏程1, 秦浩东2, 张颖1, 张珂1   

  1. （1.常州大学 安全科学与工程学院，江苏 常州 213164；2.中煤科工集团重庆研究院有限公司，重庆 400039）
  • 收稿日期:2024-05-26 修回日期:2024-09-01 出版日期:2025-04-15 发布日期:2025-04-15
  • 作者简介:赵鹏程，常州大学安全科学与工程学院，讲师，主要研究方向为危险作业人员行为智能防控技术、油气装备失效分析与完整性管理，江苏省常州市武进区滆湖中路21号，213164，zhaopengcheng0822@126.com。
  • 基金资助:
    中国石油天然气股份有限公司-常州大学创新联合体科技合作项目（KYZ22020129）

A detection method of safety equipment for personnel in fire operation under deep learning

Zhao Pengcheng1, Qin Haodong2, Zhang Ying1, Zhang Ke1   

  1. (1. School of Safety Science and Engineering, Changzhou University, Changzhou Jiangsu 213164, China; 2. CCTEG Chongqing Research Institute, Chongqing 400039, China)
  • Received:2024-05-26 Revised:2024-09-01 Online:2025-04-15 Published:2025-04-15

摘要： 安全防护装备是保障作业人员安全的重要设施，为提高人员安全装备识别精度，解决施工现场动火作业智能安全管理问题，提出一种深度学习下动火作业人员安全装备检测方法。通过在YOLOv5网络中引入CBAM注意力机制和SPD-Conv模块，优化主干特征提取网络对目标的特征提取能力，建立数据集训练网络以构建动火作业人员安全装备智能检测模型。结果表明，改进后的YOLOv5网络模型对4类安全装备的平均识别精度为92.9%，较原始网络提升了8.8%。该方法对动火作业现场人员安全装备检测具有较高的识别精度，能有效促进施工现场的智能安全管理。

关键词: 动火作业, 深度学习, 安全装备, YOLOv5网络

Abstract: Safety protection equipment is an important facility to protect the safety of operators. In order to improve the precision of personnel safety equipment identification and solve the problem of intelligent safety management of fire operations on construction sites. A detection method of safety equipment for personnel in fire operation under deep learning was proposed in this paper. The YOLOv5 network was enhanced by incorporating the CBAM attention mechanism and the SPD Conv module. The feature extraction ability of the network was improved by optimization. A dataset training network was established to develop an intelligent detection model for identifying safety equipment of pyrotechnic operators. The results show that the improved YOLOv5 network model has an average recognition accuracy of 92.9% for the four types of safety equipment, which is an 8.8% improvement over the original network. The method exhibits high recognition accuracy in detecting safety equipment at fire operation sites. It can effectively promote intelligent safety management on construction sites.

Key words: fire operation, deep learning, safety equipment, YOLOv5 network