Study On Visual Attention Patterns And Driving Behavior Characteristics Of Driving Microsleep

With the progress of current social technology and the increasing development of market economy,the number of vehicles is increasing,and a large number of accidents occur,which not only brings great economic losses to the national transportation industry,but also brings many troubles to the road traffic participants.Fatigue driving is one of the direct factors causing these traffic accidents.In-depth research on driver fatigue driving is of great practical significance to protect the property and life safety of traffic participants.The breakthrough of technology must think from a new perspective.By studying the phenomenon of microsleep in the process of drivers from waking to sleeping,it will help to reduce the generation of fatigue driving and reduce such traffic accidents from the root.Based on the driving microsleep experiment,this paper extracts the visual attention pattern and driving behavior characteristics,analyzes their laws,and extracts the visual attention pattern characteristics and driving behavior characteristic parameters.Select the data fusion theory which is consistent with driving microsleep,and establish the driving microsleep BP neural network to identify microsleep with high accuracy.The main research contents of this paper are as follows:(1)Based on the performance of driving microsleep,a specific experimental method to study the characteristics of driving microsleep is designed;The minimum number of drivers is calculated based on the standard normal distribution statistics,standard deviation and maximum error;Thirty driving microsleep experiments were carried out with the driving simulator of Southwest Jiaotong University.The driving behavior data,subjective questionnaire data and eye movement data were collected at the same time;The validity of driving simulation experiment is analyzed according to the questionnaire results.(2)The method of manual calibration is used to recognize the microsleep phenomenon in the process of driving;According to the attention mode and driving behavior characteristics of driving microsleep data,a series of evaluation indexes of driving microsleep characteristics are proposed,including: average fixation frequency,average pupil diameter,average video scanning rate,average blinking frequency,cumulative space-time area of overspeed,speed change rate,standard deviation of lane offset,number of crossing lines,steering wheel reversal rate and alternating use rate of accelerator and brake.(3)The collected data are processed,the proposed driving microsleep index is calculated,the driving microsleep is quantified,and the data obtained by the index are analyzed: speed,pupil diameter,lane offset,etc.Determine the value range of different index levels,establish the membership matrix,and use the entropy weight method to determine the weight of driving microsleep index,so as to construct the fuzzy comprehensive evaluation model of driving microsleep characteristics,and use this model to calculate the scores under different driving microsleep quantities.(4)The commonly used data fusion algorithms are sorted out,and the model of driving microsleep recognition is determined.Using correlation analysis,the indexes with strong correlation are eliminated,the key parameters of the model such as input layer,output layer,hidden layer and the number of nodes are determined in turn,and the driving microsleep BP neural network model is established.The driving microsleep experimental samples are randomly selected to verify the model,and the accuracy is as high as 83.25%.Therefore,the model can be used for driving microsleep identification,which provides new ideas and methods for fatigue driving research,and also realizes the innovation of driving microsleep identification technology.