A method for access control and monitoring compliance with safety regulations in energy facilities of enterprises based on a conveyor neural network model

The purpose of research is improving the quality of monitoring safety violations in energy facilities of enterprises through the use of a method of automated incident detection in real time, based on the conveyor application of neural network models. Methods . The article proposes a pipeline neural network model YOLO – Tesseract – YOLO, designed to solve the problem of automated access control and monitoring compliance with safety regulations in real time at energy facilities of enterprises. A method for access control and monitoring compliance with safety regulations at energy facilities of enterprises is proposed, consisting in the pipeline application of neural network models YOLOv8 and Tesseract-OCR using morphological image processing, allowing to classify a group of electrical safety clearances based on recognized patterns in an employee's ID card and detect safety violations when working with electrical installations in real time. Results . A number of experiments were conducted, during which error matrices were obtained, which made it possible to evaluate the classification quality of the pipeline neural network model using such metrics as Recall, Precision and F1-measure, the metric values were presented for all classes. The value of the F1-measure metric for the YOLO1 neural network model used to evaluate the overall efficiency, equal to 0.98, indicates a balanced relationship between the accuracy and recall of the model. The value of the F1-measure metric for the YOLO2 neural network model equal to 0.73 indicates acceptable results of the model for solving the classification problem in real time, but indicates the need to refine this part of the pipeline neural network model to improve the overall efficiency. Conclusion . The results obtained during the study indicate an acceptable quality of the pipeline neural network model in solving the problem of automated access control and monitoring compliance with safety regulations in real time. Keywords: monitoring, safety engineering, access control, energy facilities of enterprises, pipeline neural network model, Tesseract-OCR, YOLOv8, classification.