Research On Duty Personnel Authentication And Fatigue Recognition Based On Image Processing

With the rise of artificial intelligence research,face recognition and fatigue detection technologies have been applied in all walks of life,which have guaranteed the convenience of people's life.Nowadays,the staff members in some important places such as bank offices,railway operation rooms,fire monitoring rooms and intensive care units,need to stare at video image monitors all the time.Serious accidents caused by fatigue,drowsiness,or unauthorized departure of duty are not rare.Therefore,image processing technology is of great theoretical and practical value to improve the work quality of the staff.This paper studies the video-based,non-contact,real-time identity identification and fatigue detection of watchmen.First,video of watchmen is collected with the help of the camera in the duty room,and the face detection algorithm of multi-task convolutional neural network is used to quickly locate the face area and accurately locate the eyes and mouth.Then the face recognition algorithm based on convolutional neural network is used to extract face features,and the classifier is used to recognize faces.Secondly,the fatigue parameters related to eyes and mouth are extracted and positioned to judge the fatigue status of the watchman.The main work is as follows:(1)It introduces the research background and significance of personnel identification and fatigue detection system,and summarizes relevant literatures at home and abroad.In the first chapter,the research status of face detection,face recognition and fatigue recognition at home and abroad are summarized.The application of convolutional neural network in face detection and face recognition is discussed.(2)Based on the classical face recognition algorithm framework vgg-16,the network structure and loss function of vgg-16 are improved accordingly.In terms of structure,several convolution layers and a full connection layer are reduced to improve the efficiency.As for the loss function,the original softmax loss function is added with the center loss function to ensure the accuracy of face recognition classification.(3)Based on the analysis of eye state recognition method,the conventional techniques of fatigue recognition are discussed and compared.The method of combining PERCLOS value of eye,blink frequency and PMRCLOS value of mouth is used to realize the identification and early warning of fatigue.The simulation analysis shows that this method is feasible and effective.(4)The application system of face authentication and fatigue detection is designed,and the recognition function of the system is verified through the small face image data set collected in the laboratory,and the common factors affecting face recognition are tested,including illumination intensity,face posture and face occlusion.At the same time,the real-time video experiment is used to test and analyze the fatigue state,which is proved to be workable in tests.Based on image processing,the identification right and real-time fatigue status monitoring technology of the watchman can effectively identify the identity of the watchman and judge whether the watchman enters the fatigue state in time,so as to improve the work efficiency of the watchman.Meanwhile,it provides a positive way to solve the hidden security problems caused by the working status of the watchman.