基于YOLOv5的视觉风险评估方法:以高校实验室为例  

Visual Risk Assessment Method Based on YOLOv5: A Case Study of University Laboratories

作　　者：周贤伟 郭晨[1,2] 覃贞鑫 ZHOU Xianwei;GUO Chen;QIN Zhenxin(School of Safety Science and Emergency Management,Wuhan University of Technology,Wuhan 430070,China;不详)

机构地区：[1]武汉理工大学安全科学与应急管理学院,湖北武汉430070 [2]武汉理工大学中国应急管理研究中心,湖北武汉430070 [3]诺丁汉特伦特大学,英国诺丁汉NG14FQ

出　　处：《武汉理工大学学报（信息与管理工程版）》2022年第3期383-388,共6页Journal of Wuhan University of Technology：Information & Management Engineering

基　　金：国家社科基金重大项目(21&ZD127);国家文化和旅游科技创新工程项目(20211g0085)。

摘　　要：针对目前实验室事故多发的问题,选取频率较高、损失最大的爆炸事故作为研究对象,提出基于YOLOv5算法的视觉风险评估模型,使用目标检测手段提高实验室安全监控能力。通过统计学者对实验室爆炸事故分析得到的影响因素,确定5种安全隐患因素。对YOLOv5算法采用数据增强、迁移学习等手段实现实验室安全隐患的高精度检测,检测精度为0.987。最后建立风险评估规则,设计人工评估与模型评估的对比实验,结果表明:模型评估方法具有更高的检测精度,更好的稳定性,以及更快的检测速度。模型评估方法可以使管理人员实时获知实验室风险等级,进而提高实验室安全管理水平。In view of the frequent occurrence of laboratory accidents, the explosion accident with high frequency and maximum loss are selected as the research object. A visual risk assessment model based on YOLOv5 was proposed to improve the safety monitoring ability of the laboratory by object detection. Through the statistical analysis of the influencing factors of laboratory explosion accidents, five potential safety hazard factors are determined in this paper. In this paper, the YOLOv5 algorithm is used to achieve high precis-ion detection of laboratory security risks by means of data enhancement and transfer learning, and the detection accuracy is 0.987. Finally, the risk assessment rules are established, and the comparative experiments of manual assessment and model assessment are designed. The experimental results show that the model assessment method has higher detection accuracy, better stability and faster detection speed. The model assessment method allows managers to be informed of laboratory risk levels in real time, thus improving laboratory safety management.

关 键 词：实验室爆炸事故 风险评估 YOLOv5算法 迁移学习 目标检测 

分 类 号：X932[环境科学与工程—安全科学]