Research On Face Mask Wearing Detection And Standard Wearing Recognition Algorithm

Wearing masks has proven to be the most effective approach of epidemic prevention as the COVID-19 epidemic rages across the globe now.Wearing a face mask and following the standard code in public places can help reduce the spread of viruses and avoid cross-infection.How to detect if a person is wearing a mask and how to determine if the mask is properly worn have become a topic of increasing interest in recent years.We are now in the third year of this worldwide pandemic.From Delta to Omicron,the virus is still evolving and posing an ongoing threat to the well-being of the humanity.For the benefit of people’s lives,health,and safety,China’s epidemic prevention and control efforts are gradually returning to normalcy in the face of a complicated and severe epidemic situation.The government has repeatedly asked people to wear masks scientifically in crowded places like public areas.There are,however,some people who do not wear masks or even don’t wear them because of the troubles and discomfort of wearing masks.They are lucky that they have not caught the virus,as their actions greatly increase the risk of virus transmission.Therefore,it is essential to check and remind pedestrians to wear masks in public places.Manual inspection is ineffective,slow,and prone to false detections and missed detections.The use of intelligent devices to automatically detect the wearing of face masks has become an inevitable trend.This paper investigates the mask wearing detection and standard wearing recognition algorithm,including the detection of whether a multi-face target wears a mask in the natural scene and the recognition of whether a single-face target wears a mask in the active scene,adopting the deep learning method.The work contents are as follows:(1)The research efforts and accomplishments of face mask wear detection and standard wear recognition algorithms at home and abroad are analyzed in depth and systematically.The advantages and disadvantages,primary solutions,and model improvement strategies of different types of algorithms in active and passive application scenarios are summarized,along with the existing issues of face mask wear detection and the dearth of standard wear recognition research.(2)YOLOv4-tiny’s improved attention feature fusion is proposed as a mask-wearing detection model to address the problem of missed detection and false detection in the multi-mask face target detection scene.By incorporating the attention mechanism module ECA-Block and path aggregation network,the accuracy of the model is enhanced,the real-time requirement is met,and the optimal detection effect is achieved.The performance on the value of the data set is 91.90% m AP and 30 FPS respectively,which proves the effectiveness of the model.(3)A novel lightweight two-stage mask standard wearing recognition model is proposed to address the standard wearing recognition issue in the application scenario of a single mask face target.When combined with Ghost Net,face detection network MTCNN,and lightweight convolution network Ghost Net are employed to classify the wearing situation.The whole algorithm is realized by programming,and the real-time test demonstrates the validity of the model.