Production Workshop Safety Management And Control System Based On Deep Learning

Heavy industry is the backbone of our country’s national economy.With the development of society and the advancement of science and technology,our country has higher requirements for safety management and control of heavy industry production.At present,the system functions on the market are relatively simple and tend to be business management,and there are problems of low intelligence and low security.The development of the production workshop safety management and control system based on deep learning on the one hand enriches the functional modules of the system,and on the other hand realizes the helmet-wearing detection algorithm,which is conducive to improving the system’s intelligence and safety management and control capabilities,and has strong practical significance.This thesis builds a production workshop safety management and control system based on the B/S architecture.The system uses Spring Boot and Vue to separate the front and back ends,and uses My SQL and Redis to build a distributed database.The system includes six functional modules,namely personnel management module,visitor management module,alarm management module,video inspection module,equipment management module,and dangerous operation management module.The personnel management module manages the information of employees,VIPs,and blacklisted personnel.The visitor management module controls the application,approval,and other processes for visitors entering and leaving the company site.The alarm management module implements the helmet-wearing detection algorithm,triggers device alarms by setting analysis tasks and alarm rules,and realizes real-time alarms.The video inspection module can formulate inspection plans to inspect the production process and report inspection records.The device management module manages company devices.The hazardous operation management module manages the application,execution,change,and closure of hazardous operations.During the process of hazardous operations,inspectors can inspect the working conditions of different areas in the video inspection module.In the alarm management module,this thesis implements a helmet-wearing detection algorithm based on YOLOv5.This algorithm improves the CSP module of the original Yolov5 algorithm in two aspects,thereby improving the detection ability of the algorithm to small targets.First of all,Dense Net and CSP are integrated,and the feature of dense connection of Dense Block is used to maximize the ability of the network to extract feature information to improve the ability to detect small objects.Secondly,the SE-Net attention mechanism is added after the Conv of CSP,so that the network pays more attention to the feature information of small targets,thereby improving the ability to extract feature information of small targets.Through the simulation test,it is concluded that the accuracy of the improved algorithm is 96.2%.The comparative experimental results show that the m AP of the algorithm in this thesis is improved compared with the original YOLOv5 network on the self-made helmet wearing detection dataset or the VOC2028 public dataset.The helmet-wearing detection algorithm based on YOLOv5 proposed in this thesis is more effective for helmet detection.The system has been put into use in many companies,providing convenient services for companies,saving human,material,and financial resources,and effectively reducing the incidence of safety accidents in the production workshop.