Driver’s Unsafe Driving Detection Method Based On Deep Learning

In order to reduce car accidents caused by unsafe driving,this paper designs a driver unsafe driving detection method based on deep learning.This method is an automated detection method based on deep learning,which is divided into four parts: face recognition,mask recognition,human posture detection,and fatigue detection.The camera is used to collect video frames in real time,and the lightweight network is used for face recognition and mask detection,and then the human body gesture recognition is performed to determine whether the driver answers the phone,which reduces the amount of model calculation and improves the detection accuracy.The traditional method of detecting fatigue driving is to identify the driver’s facial signals to determine whether to yawn and close his eyes.Because the driver has been wearing a mask during the epidemic,the facial signals are easily disturbed,and the person’s EEG,eye movement,and EMG signals when fatigued,Skin electrical signal,body temperature,heart rate,blood pressure,etc.will fluctuate,so it is proposed to use the cerebellar neural network to fuse the various physiological signals of the driver to determine the fatigue level.Compared with the experimental results of a single signal,the result obtained by fusion of multiple signals is closer to the true value.In the face recognition module,the MTCNN network model is used for face frame positioning,the Face-Net network model is used for face key point comparison,and the Mobile Net V2 network model is used for mask detection.In the attitude detection module,the open-pose library is used to locate the key points of the driver’s human skeleton.The Mobile Net V2 network model is used to determine whether there is a mobile phone in the driver’s left and right ear area.If there is a mobile phone,the driver will be judged to call while driving.In the fatigue detection module,use EEG headsets to detect EEG signals,eye trackers to detect eye movement signals,multi-lead physiometer to detect signals such as myoelectricity,skin electricity,skin temperature,heart rate,etc.,and use cerebellar neural network to fuse multiple physiological signals to determine Driver fatigue level.The driver’s unsafe driving detection method based on deep learning can recognize the driver’s face in real time through the on-board camera,detect whether the driver is wearing a mask through mask recognition,and obtain the fatigue level through the integration of multiple signals with physiological signals,starting from the driver himself,and reducing traffic accidents happened.