Distraction Identification And Risk Analysis Of Using Mobile Phone While Driving

Driving distraction is a key factor effecting driving safety.In order to effectively discriminate the distraction state during driving process,this paper collected and analyzed the eye movement and vehicle operation data in different driving states by FORUM 8 driving simulation experiment platform and D-Lab driving behavior monitoring system.Then,we established distraction identification model and analyzed the rear-end collision risk.The specific research contents are as follows:(1)An experimental scheme for acquiring visual and vehicle running characteristics was proposed.A total of 12 drivers were selected to conduct the normal driving and driving distraction test,so as to obtain eye movement and vehicle operation data.The method of extracting eye movement data was proposed,and the original coordinates of gaze points were preprocessed.(2)The characteristics of eye movement and vehicle operation in different driving states were analyzed.K-means clustering algorithm was used to cluster the coordinates of the fixation points.Drivers' visual area of interest was divided into 7 parts according to the clustering results.The data of distribution ratio of gaze points,frequencies and duration of gazing at front window AOIs,percentage of gazing away from the front road,search breadth in horizontal and vertical directions,saccade size,saccade duration,saccade speed,standard deviation of speed,mean acceleration and standard deviation of acceleration were extracted from the experiment results.Then the difference analysis between these parameters were realized using the F test method.(3)A driver distraction identification model was constructed.According to the results of difference analysis,there were 6 parameters selected as input variables of the model.The parameters of the support vector machine kernel function were optimized by genetic algorithm.The driving state was identified using the radial basis function.The classification accuracy of the model is 93.49%,which could accurately identify driver distraction state.(4)The risk of rear-end collision in driving distraction state were analyzed.This paper analyzed the process of rear-end collision and methods of accident prediction.Based on the conflict circle model,critical conflict area model and driver reaction time,the number and severity of rear-end conflicts in different driving states were judged.According to K-means clustering algorithm,the risk of rear-end conflict was divided into four levels: low-risk zone,medium-risk zone,medium-high-risk zone and high-risk zone.The result shows that the average reaction time of mobile phone conversation is 0.16 s later than that of normal driving and the average reaction time of texting is 0.3 s later;rear-end conflict are more frequent in distracted driving states,and the peak risk level is about twice as high as that of normal driving;the rear-end conflict risk of texting is the highest.