Design And Implementation Of Mask Wearing Detection System Based On YOLOx

In early 2020,covid-19 outbreak was reported.According to the latest data released by WHO in March 29,2022,the total number of new crown confirmed cases exceeded480 million,and over 6 million 100 thousand people died.The recurrence of the epidemic has brought great challenges to China’s public health system.According to the management theory,the essence of epidemic prevention and control is the phenomenon of resource run caused by the outbreak of instantaneous demand,which is also the wicked problem in management.Blocking the path of virus air transmission can effectively reduce the possibility of large-scale outbreak.Therefore,in places with relatively dense flow of people,it is particularly important to ensure that personnel wear masks and keep a safe distance.Taking the target detection algorithm as the technical core,this thesis improves the YOLOx algorithm for the face wearing mask detection scene.The research contents of this thesis mainly include:For the scene of face skin color detection,the YOLOx algorithm is improved.For the indoor face wearing mask detection scene,it is proposed to add channel attention mechanism on the basis of YOLOx to improve the weight of face image in target detection,so as to improve the accuracy of face target detection.In order to avoid the problem that the gradient disappears or it is difficult to accurately reflect the size and shape of the rectangular box in the calculation of the loss function,this thesis uses the ciou（complete intersection over union）loss function to replace the original IOU（intersection over union）loss function.Experiments show that the accuracy of the improved YOLOx algorithm in face detection scene is higher than that of the improved YOLOx algorithm.Aiming at the problem that it is difficult to detect whether the mask meets the standardized standard in the process of wearing,this thesis proposes a secondary target detection method to realize the recognition function of non-standard wearing mask.Considering that in the actual scene,some personnel will expose their nostrils,this kind of non-standard wearing of masks will make the protective effect of masks useless.In order to realize the recognition function of nonstandard wearing masks,this thesis proposes a secondary target detection method.Compared with the direct target detection of the nose tip or mouth with a small area,the classification of the whole picture without standard mask is more conducive to improve the accuracy of detection.In view of the lack of training set of nonstandard wearing masks,this thesis makes a self-made data set by superimposing mask images above the mouth position,and synthesizes a variety of nonstandard wearing mask images as training set and verification set.Aiming at the problem that the position of the mouth can not be determined,this thesis makes the data set by manual marking,trains it in the YOLOx network,and obtains the eye and mouth position detection model to determine the position of the mask map and the rotation angle of the mask.Experiments show that the standard wearing mask recognition network can effectively identify the situation of non-standard wearing mask.Aiming at the application problems in the actual scene,this thesis designs and implements a mask wearing detection system based on improved YOLOx.Aiming at the problem of safe distance detection,the system determines the position of queuing personnel through infrared sensors.Aiming at the problem of mask wearing,the mask wearing situation is recognized through the mask wearing detection model.And the system can remind the queuing personnel to wear masks and keep a distance of one meter offline.Managers can monitor the queuing site through the system.The system consists of webcam,infrared sensor,local computer,main control chip,electronic display screen,loudspeaker,buzzer and other hardware.The system obtains the on-site live information through the webcam,obtains the location information of the on-site queuing personnel through the infrared sensor,identifies the image information through the local computer,judges whether the on-site personnel wear masks regularly,and synchronizes the information to the web page and mobile terminal through the cloud platform.If it is detected that the mask is not worn,the mask is not worn in a standard manner,or the distance between people is too close,the system will call the speaker,electronic display screen and buzzer to broadcast the corresponding audio and picture through the main control chip to remind the on-site queuing personnel to wear the mask in a standard manner or keep a safe distance.Figure[40] table[16] reference[75]...