Intelligent Identification of Personnel off Duty Based on Improved YOLOv5

As one of the unsafe behaviors in the safety production process, the absence of operators has laid a major safety hazard for the safety production of the enterprise. Once a problem occurs in production, it is easy to cause serious casualties and property losses due to the inability to warn and handle it in time. This paper proposes an intelligent identification method based on improved YOLOv5 for the absence control of the SCRs (security control rooms). By introducing BotNet into the YOLOv5 architecture, the neural network can improve the feature extraction ability while paying more attention to the key features of the target area and improve the model precision by 3%. At the same time, the model is trained through data enhancement strategies such as HSV and MOSAIC techniques to reduce the similarity between data and hence gained the diversity of data. The model is experimentally verified by the enterprise production data obtained from real scenarios. The results show that the intelligent identification method based on improved YOLOv5 has significantly boosted the performance in identification accuracy, specificity and F1 score.