摘要
不安全行为是导致化学实验室事故的重要原因之一,针对化学实验室人员不安全行为模式识别的研究具有现实意义。本文首先定义了5种化学实验室人员不安全行为模式,进而根据某高校化学实验室实际场景,构建了不安全行为模式视频数据集,并建立基于时域分割网络(TSN)的化学实验室人员不安全行为模式识别模型,在模型中增加dropout层,最后使用SoftMax分类器对不安全行为进行识别。结果表明:本文模型可以准确识别本文定义的化学实验室不安全行为,召回率、精确率大于0.9,F1指数大于0.95,预期可以为化学实验室人员不安全行为的预警和风险防控提供技术支持。
Unsafe behavior is one of the important causes of accidents in chemical laboratories,so research on the identification of unsafe behavior pattern is extremely necessary.In this paper,by means of investigation of chemical laboratory regulations,five typical unsafe behavior patterns are defined,which are drinking water,eating food,playing games with mobile phones,sleeping and smelling reagents.Then a video dataset containing unsafe behavior patterns in the actual scene of chemical laboratory in a university is constructed.Finally,the pattern recognition model of unsafe behavior in chemical laboratory based on the temporal segment network(TSN)is established.A dropout layer is added to the model to prevent data overfitting,and the above five unsafe behavior patterns are identified.The results show that the pattern recognition model of unsafe behavior in chemical laboratory based on the temporal segment network can accurately identify the five unsafe behavior patterns defined in this paper.The average recall rate and precision for five unsafe behavior patterns on the testing sets are higher than 0.9,and the average F1-score is higher than 0.95.The results show that the model can accurately identify the unsafe behavior mode of chemical laboratories defined in this paper.The results of this research are expected to provide technical support for the early warning and risk prevention of unsafe behaviors in chemical laboratory.
作者
黄振邦
HUANG Zhenbang(School of Information Network Security,People's Public Security University of China,Beijing 100038,China;Key Laboratory of Security Prevention and Risk Assessment,Ministry of Public Security,Beijing 100038,China;Public Security Behavioral Science Laboratory,People's Public Security University of China,Beijing 100038,China)
出处
《智能计算机与应用》
2022年第2期99-104,共6页
Intelligent Computer and Applications
关键词
不安全行为
化学实验室
模式识别
时域分割网络
unsafe behavior
chemical laboratory
pattern recognition
temporal segment network
作者简介
黄振邦(1996-),男,硕士研究生,主要研究方向:深度学习应用。
