| Food health and safety is an important factor affecting public health and social welfare.Food health and safety has become the focus of people’s discussion after dinner.Although people’s attention has been increasing year by year,the health accidents caused by diarrhea and other diseases caused by eating unhealthy food still occur continuously and increase year by year.There are many inspection items in food safety monitoring work,which need more supervisors and inspectors.With the continuous development of advanced management concepts and information technology,food safety supervision and management and kitchen video monitoring system construction provide food supervision and management departments and relevant agencies with relatively transparent and intelligent information basic data of food safety supervision,so as to achieve a new level of food safety supervision and management.In the process of intelligent monitoring and detection of kitchen staff’s health situation,we will inevitably encounter the problem of face chef’s hat and mask blocking,and how to solve the problem of face recognition with blocking is the key of this study.The problem caused by face occlusion is not only that some organs on the face are occluded by unknown objects,but also the most important is that it will make the key features of the face unable to be extracted completely and accurately,resulting in the loss of image information.In this paper,the intelligent monitoring algorithm of kitchen staff based on machine learning is studied.Firstly,face image information is collected at the door of kitchen to recognize the situation of wearing chef’s hat and mask,and enter when wearing complete.During the working period,real-time detection shall be carried out for allpersonnel to see if they remove the chef’s hat and mask in violation of regulations,and warning shall be given.The main research contents and conclusions are as follows:(1)First,we use SVM to classify the face images of chefs wearing hats and masks.Then we divide the face images into four categories: all chefs wearing hats and masks,all chefs wearing hats and masks,all chefs wearing masks and all chefs wearing masks without masks.Then,the face image is divided into 25 sub image blocks by using ulbp operator to reduce the dimension of the face image,and the one-dimensional information entropy weight of each sub image block is weighted to the feature extracted by the sub image block.Finally,we use the depth confidence network(DBN)to classify the occluded faces and detect and recognize the kitchen staff information.The average accuracy of face recognition method based on EWLBPDBN is 96.47%.(2)In this paper,the video target detection algorithm based on the cuckoo search algorithm and the Gaussian mixture model(CSGMM)is used to detect the video during the work in the kitchen.In this method,the cuckoo search algorithm is introduced into the Gaussian mixture background modeling to find the appropriate learning rate and foreground threshold.Compared with the traditional Gaussian mixture model,it is more detailed and accurate in the detection of moving objects,and achieves a better detection effect.It not only improves the integrity of moving object detection,but also enhances the accuracy of background description.(3)The mask detection algorithm in the working area can detect the situation that the staff take off the mask illegally in real time.The CSGMM algorithm in(2)is used to improve the fast-rcnn convolutional neural network to remove the interference of complex background and improve the detection accuracy of fast-rcnn.The algorithm used in this paper has achieved high accuracy and recall rate of mask face detection.Compared with other detection algorithms,it has excellent performance and basically meets the needs of kitchen video detection. |
PDF Full Text Request