A Study On The Safety Monitoring Of Machine Tool Operators Based On Facial Status

In recent years,with the continuous development of China’s economy,China’s manufacturing level is also rising,and gradually become a pillar industry of rapid economic development,but customers present a variety of product needs,the precision of the parts more and more demanding,so the requirements of the machine tool is also increasing,at the same time the development of the industry also makes the pressure on the operator is growing,the operator is more and more likely to produce fatigue,some fatigue Some fatigue behaviour can lead to major machine tool accidents,causing great damage to the operator’s life and property.Therefore,it is very important that fatigue detection is carried out quickly and accurately,so it is important to study fatigue detection technology in order to reduce operating accidents and establish a safe operating environment.In this paper,based on a combination of domestic and international fatigue detection,we conduct an in-depth study,based on a variety of computer technology to locate the face area,through the now more commonly used PERCLOS and PMOT multi-indicator fusion judgment method to determine the fatigue state,and finally by the SSD network to monitor the operator’s work status in real time.The main research elements are as follows.(1)Research on operator face detection algorithm.For the face region detection problem,the fast face detection algorithm of Retinaface+Facenet is used.The Retinafac algorithm has two networks that can be used as the backbone feature extraction,namely Mobilenet V1-0.25 and Resnet,while the Mobile Net model is a Google proposed for embedded devices The Mobile Net model is a lightweight Depthwise separable convolution network proposed by Google for embedded devices,which divides the full convolutional operation into two steps,thus greatly reducing the number of parameters and making it ideal for deployment in embedded devices at a size of 1.68 M.space.The model requires only a very small amount of processing of the image.(2)Operator fatigue detection algorithm research.For the task of real-time operator fatigue detection,a fatigue detection algorithm based on single-target detection(SSD)is designed,which extracts multiple facial fatigue features such as eyes and mouth to jointly determine the degree of fatigue from multiple angles.The fatigue recognition method in this paper achieves an average accuracy rate of 98.7% for the eyes and 99.2%for the mouth,and meets the demand for real-time fatigue recognition on the embedded platform.(3)Fatigue detection system based on the embedded platform.Based on the fatigue detection algorithm proposed in this paper,the fatigue detection system is embedded in the dw-79 embedded platform,and the operator’s working status is shown in real time through the camera and the UI interface.The system has a simple and clear interface,with a full range of functional buttons,and can also give alerts while fatigue is being detected.