摘要： 针对因人的不安全行为导致化学实验室安全事故的预防问题，提出了基于三维卷积网络(Convolutional 3D，C3D)的化学实验室人员不安全行为模式识别方法。定义了5种典型的人员不安全行为模式，以某高校化学实验室作为研究区域，构建了人员不安全行为模式数据集，建立了基于C3D的人员不安全行为模式识别模型，并对模型在不同应用场景下的性能进行了评估。结果表明：在特定人员于特定实验场景中出现上述5种不安全行为时，该模型在测试集上得到的F1指数平均值均能达到97%以上；在非特定人员与非特定场景中出现不安全行为时，模型能够有效识别部分不安全行为。该研究成果预期可以为化学实验室人员不安全行为的预测、预警、预防提供技术支持。

中图分类号： TP391.4；TP181
文献标识码： A
DOI： 10.19358/j.issn.2096-5133.2022.03.012
引用格式： 刘凯源. 基于C3D的化学实验室人员不安全行为模式识别[J].信息技术与网络安全，2022，41(3)：71-77.

Patterns recognition of unsafe behavior in chemical laboratory based on C3D

Abstract： In order to prevent chemical laboratory safety accidents caused by unsafe behavior，a patterns recognition method of unsafe behavior in chemical laboratory based on 3D convolutional network(C3D) is proposed. Firstly, five typical unsafe behavior patterns are defined. Then a university chemical laboratory is used as the study area to construct a dataset containing simulated unsafe behavior records. The patterns recognition model of unsafe behavior in chemical laboratory based on C3D is finally established, and the model performances in different scenarios are validated. The results show that under the circumstance that observed unsafe behaviors are from a specific person in a specific experimental scenario, the average F1-score on test set exceeds 97%. As well as, for non-specific persons and scenarios, the model can effectively identify some unsafe behaviors. The results of this research are expected to provide technical support for prediction, early warning and prevention of unsafe behaviors in chemical laboratory.

Key words : unsafe behavior；chemical laboratory；pattern recognition；3D convolutional network