Research Of Fatigue Driving Detection System Based On Face Recognition Technology

Traffic accident occurs frequently as the number of vehicles and drivers continue to increase.Among them,the amount of accidents caused by driver fatigue has increased significantly.In order to reduce the probability of such accidents,this paper based on the face recognition technology,identifies the fatigue characteristics in the driver's individual unit,and comprehensively judges the driver's fatigue state through multi-information fusion algorithm.And make a voice alarm prompt.Firstly,the image is preprocessed by adaptive median filtering and threshold segmentation algorithm.The fast PCA target feature extraction algorithm combined with SVM classification algorithm is used to identify the driver's face,and the AdaBoost classifier is trained to locate the human face,the human eye and the human mouth region.The experiment uses the near-infrared database image batch location recognition to verify the algorithm.In order to improve the real-time performance of the system,the Mean Shift algorithm and the interframe difference method are combined to track the target motion trajectory to predict the detected feature position in the next frame image.Secondly,in terms of feature extraction,the vertical projection method is used to determine whether the driver wears glasses.Among them,the unworn glasses are identified by the width is unchanged and the height is 1/3 of the eye height as the threshold value compared with the normal state.The glasses are processed by the image and then the same operation is performed.The Canny operator is used to obtain the edge feature of the mouth,and the opening and closing state is recognized by the roundness of the mouth.The head uses the second-order HOG feature combined with the SVM regression algorithm to perform head angle recognition.The experiment verifies that the yawn and nod in the auxiliary feature are highly correlated with the judgment of the driver's fatigue.However,the single feature judgment fatigue has limitations,so further information fusion is needed.Finally,the fusion strategy of this paper is combines the three characteristic parameters of PRECLOS,closed eye duration and blink frequency of the eye at the feature layer based on the feature-weighted Bayesian algorithm.And based on the dynamic Bayesian network,the multi-information fusion of the multi-feature parameters of the eye,the characteristics of the mouth yawn feature and the head nod feature parameters in the decision-making layer is carried out.The dynamics of the dynamic Bayesian network can optimize the fatigue in the time period.Through the design experiment,by adjusting the weight of each evidence variable,distinguishes the weight of the algorithm when wearing glasses,and experiments verify that the fatigue can be more effectively determined by changing the weight.