Design And Implementation Of Fatigue Driving Detection System Based On Facial Multi-Feature Fusion

With the continuous increase in the number of vehicles and drivers in China,the number of casualties and property losses caused by traffic accidents has been rising year by year,and fatigue driving has become one of the main causes of traffic accidents.Therefore,researching a safe and reliable fatigue driving detection system has important theoretical and practical significance.Among the commonly used methods for detecting fatigue driving,those based on driver’s physiological characteristics and those based on vehicle driving characteristics have problems such as expensive equipment,susceptibility to environmental influences,and single detection indicators.In view of these problems,this paper proposes a fatigue driving detection method based on multiple facial features fusion and constructs a fatigue driving detection and supervision system based on Orange Pi 5 embedded development board,which can accurately analyze the driver’s fatigue state and improve driving safety.The specific research contents are as follows:(1)A lightweight YOLOv5s face detection network based on the SimAM nonparametric attention mechanism is designed.In order to enable the model to focus more attention on effective information areas and improve the detection accuracy of the model,the SimAM non-parametric attention mechanism is added to the tail of the YOLOv5s backbone network.The experimental results on the Wider Face dataset show that compared with the original YOLOv5s,the improved network increases the detection accuracy by 2.1%under the same parameter volume and can better complete the face detection task.(2)A PFLD face key point detection model with an auxiliary sub-network is improved.Firstly,a multi-scale feature fusion module is constructed to further enhance the model’s detection ability for large and small targets.Secondly,the Ghost Bottleneck is used to optimize the model’s backbone network,further reducing the model’s parameter volume and improving the detection speed.The experimental results on the WFLW dataset show that the improved model has faster detection speed and higher accuracy,which can meet practical use requirements.(3)A fatigue judgment method based on multiple facial features fusion is studied.Firstly,the EAR and MAR values are calculated through facial key points to extract the driver’s eye and mouth features.At the same time,the PFLD auxiliary sub-network is used to obtain head features.Then,based on the above three types of facial features,fatigue thresholds are set for blink,yawn,nodding and other states.Finally,a fatigue judgment method based on multiple facial features fusion is obtained.The experimental results on the YawDD dataset show that the detection accuracy of this method reaches 95.3%.(4)A fatigue driving detection and supervision system is designed and implemented.A fatigue driving detection system based on the Orange Pi 5 embedded development board is built,and a background supervision system is developed using Spring Boot.Through testing experiments,the fatigue driving detection and supervision system designed in this paper can relatively accurately judge fatigue driving behavior.