Recognition Method Of Unsafe Behavior In Chemical Laboratory Based On 3D-CNN

The unsafe behavior in chemical laboratory is one of the main reasons for laboratory safety accidents.The existing chemical laboratory personnel safety management methods are usually low in efficiency and high in cost,which have a significant demand for low-cost,intelligent monitoring and early warning methods.In this paper,a recognition model of unsafe behavior in chemical laboratory based on 3D-CNN is established.According to the real scene of a chemical laboratory in a university,a video dataset is constructed and the model is verified.Finally,a risk analysis model of unsafe behavior in chemical laboratory based on Bayesian network is established.The main research contents of this paper are as follows.（1）A video dataset of unsafe behavior in chemical laboratory was constructed.Taking a chemical laboratory of a university as the research area,5 typical unsafe behavior patterns in chemical laboratory（Smell Reagent,Eat Food,Drink Water,Play Mobile Phone and Sleep）were selected as identification objects through investigation and follow-up observation.8volunteers were invited to simulate unsafe behaviors in different experimental scenarios and use cameras to capture the video data.After screening and classification,a dataset containing 5028 video clips was obtained.（2）A recognition model of unsafe behavior in chemical laboratory based on 3D-CNN was constructed.Through verification,the model has achieved good results on the self-constructed dataset.For the 5 typical unsafe behaviors,the average Recall,Precision,and F1-score of the model are all higher than 98%.In addition,the pattern recognition stability of the model was further examined.The model can more accurately identify the unsafe behaviors of specific personnel in specific scenarios.For unsafe behaviors of non-specific personnel in non-specific scenarios,the model can identify partial behavior pattern（taking "Sleep" as an example,the average Recall is 99.67%,the Precision is 95.67%,and the F1-score is 97.76%）.（3）A risk analysis model of unsafe behavior in chemical laboratory based on Bayesian network was constructed.Extract feature information from three dimensions of the scene,people and behavior patterns,then use the features to build a Bayesian network.Combined with the recognition results of the pattern recognition model and expert knowledge,parameter analysis is carried out to realize the risk analysis of unsafe behavior in chemical laboratory.From the results,the unsafe behavior of personnel is the key factor leading to the increased safety risk in chemical laboratories.The results of this paper can be used in chemical laboratory scenarios,and are expected to provide technical support for the prevention and control of unsafe behavior in chemical laboratory.