Facial Fatigue Expression Recognition Based On Machine Learning

In recent years,with the continuous development of the social economy,China's industrial manufacturing level has been continuously improved,and it has gradually become the pillar of rapid economic growth.At the same time,however,the rapid development of the industry has also brought about the continuous increase of personal pressure.Industrial workers are more and more prone to fatigue,affecting the production safety and production efficiency of enterprises.Therefore,rapid and accurate fatigue testing of workers has become a top priority.Based on relevant research in the world,this thesis has carried out the following research on fatigue detection in industrial production.In this thesis,the first step is data collecting and data fusing.During the acquisition process,the monitoring device cannot capture valid facial information because the subject's spatial position and head space posture are constantly changing.In response to this problem,this article uses a multi-device monitoring fusion scheme to monitor the subject's face.First,facial feature point recognition is performed on the data collected by each device.Next,data fusion is performed using a fusion method based on the minimum decision of the rotation angle of the face.Finally,the optimal facial feature po:int data set is obtained,and a fatigue feature sample database is constructed.Next,fatigue feature analysis and modeling are performed.Firstly,extract the fatigue features based on the facial optimal feature point set,and the statistical analysis method is used to analyze and verify the validity of these features.Then,for the problem of changing shooting distance and shooting angle,a feature compensation method based on distance mapping and a feature compensation method based on Euler angle are proposed.Experiments show that these two methods can effectively reduce the impact of shooting distance and shooting angle on the feature value.Finally,the optimal kernel function is selected based on the detection rate of the single-core support vector machine,and the single-branch decision tree and the support vector machine are combined to construct the operational fatigue detection model.The experimental results show that the model has a higher detection rate than the single-core support vector machine model.At the end of this thesis,the server architecture design of the system is introduced.The system server encapsulates the above detection algorithm through the data monitoring layer,the data calculation layer and the data storage layer to realize data collection,calculation and storage.This thesis focuses on the technical solutions and key execution processes of the data computing layer and storage layer.