Design And Implementation Of Laboratory Security System Based On Target Detection

With the improvement of living standards,enterprises and individuals are increasingly aware of security,security monitoring systems have been applied to a number of practical scenarios close to life,such as factory workshops,office buildings and important places.In recent years,the state has increased investment funds for university research sites,and laboratory development has become an important part of university construction,along with frequent safety accidents in university laboratories,and laboratory safety-related issues are increasingly prominent.In order to reduce the possibility of experimental personnel being injured in accidents and the loss of the laboratory,a full range of real-time monitoring of laboratory area personnel is required.Analysis of the current security problems in the laboratory,the design of an intelligent laboratory security system is of great significance.Security system is a complex system engineering,the research in this paper mainly includes: identity verification based on face recognition and lab personnel safety behavior specification detection,with mask wearing,smoking and limited area intrusion detection as examples.The research provides a new way of thinking,method and practical example for laboratory security.The main work and innovations in this paper are summarized as follows.(1)In the detection process of non-standard behaviors,an improved method based on YOLOv5 s model is proposed to solve the problem of low detection accuracy of multi-class targets.First,this paper enhanced the small target image of the data set to solve the problem of high missing rate and low detection accuracy caused by the small target size in the detection task.After the enhancement,the small target recall rate increased by 16%.Secondly,the anchor frame strategy is compared and analyzed,and the anchor frame strategy which is more suitable for this data set is selected.CBAM attention mechanism module was added to the YOLOv5 network structure to improve the network model’s attention to small targets,and m AP was increased by 2%.Finally,the sparrow search algorithm was used to find the optimal solution for the initial parameters of the network model,and its m AP was improved by 4.4%.The experimental results show that the detection accuracy of the new model is improved obviously,while the missing rate is reduced,and the detection effect is improved significantly.This shows that the improved model has better robustness and stability,and can adapt to the detection requirements of various complex scenarios.(2)A face recognition detection method based on vivisection detection is proposed for personal identity verification,so as to solve the problem of using false faces to deceive in the current face recognition process.In this paper,the YOLOv5 s algorithm is used to realize facial region detection under complex background.The network model is trained by public face detection data set,and the cosine similarity value is calculated by Arc Face algorithm to match the face information.The in-vivo detection mechanism is introduced,and the face recognition accuracy reaches 89.3% by testing on the self-made face database.The accuracy of face recognition based on vivisection detection is 69.3%.(3)Integrate the identity verification function of experimental personnel based on face recognition and the personnel behavior specification detection function based on YOLOv5 s into the system to build a laboratory security system based on target detection.The visualization interface of the security system is also designed to make it more convenient to observe the detection effect of each module intuitively.Finally,after testing,the system can operate stably.This paper delves into deep learning related algorithms,and analyzes and compares various aspects of the algorithm m AP,recall and accuracy indexes during the experiments.Through a comprehensive analysis of the experimental results,YOLOv5 target detection algorithm performs well during the above three functional tests,and the algorithm is feasible for building laboratory security systems.