Research On The Characteristics Of Driving Distraction In High-complexity Scenes And The Discrimination Model Of Driving Distraction

In the road traffic system composed of people,vehicles,and road environment,human factors account for the largest proportion of traffic accidents.Among them,inattention due to driver distraction is the most common cause of traffic accidents.Therefore,it is urgent to conduct research on driving distraction performance and distraction detection.Driving distractions are mainly divided into two categories: visual distractions and cognitive distractions.According to the classification of driving distractions,this article mainly completes the following aspects:(1)Based on Prescan software,it builds a simulation experiment scene of driving distraction in a high-complexity scene.Aiming at the problems of single simulation scene in existing research experiments and insufficient accuracy of the distraction discrimination model,the research direction of this article is proposed.Combined with the definition of driving distraction,distraction is set to replace subtasks,and a highcomplexity scene driving distraction scenario is proposed.It provides solutions to problems such as start-stop time synchronization,time step synchronization of data collection,running time synchronization of Prescan and simulink coupling simulation program,and time calibration of distracted sub-tasks,and standardizes the data to identify distractions for subsequent driving.The construction of the model provides data support.(2)Obtaining visual distraction and cognitive distraction data of different difficulty in three types of scenes(straight section,stop intersection and turn section).Using the high-complexity road as the experimental scene,4 sets of variables(gender grouping-between-group variables,age grouping-between-group variables,visual/cognitive tasks-in-group variables and complex grouping-in-group variables)were set for driving simulation.Analyzing the impact of drivers(gender,age)performing visual and cognitive subtasks of different difficulty on different road sections(going straight,stopping and turning)on the performance of driving distraction characteristics,and summarizing the significance of the indicators of each distraction characteristic,as distracted features provide a basis for the principal component sorting.(3)Based on the improved SVM-Adaboost algorithm,a driving distraction discrimination model is constructed.Based on principal component analysis,the key features that characterize driving distraction are extracted,and 7 feature indicators such as fixation time ratio,fixation point coordinates,vehicle speed,acceleration,steering wheel angle,braking force,and accelerator pedal opening degree are extracted,and the integration of two machine learning algorithms is integrated.A driving distraction discrimination model based on SVM-Adaboost is constructed,and the driving distraction discrimination model is combined with the driving distraction dangerous state discrimination method to evaluate the early warning of the driving distraction dangerous state.The research results show that the accuracy of the improved SVM-Adaboost driving distraction discrimination model is significantly improved,and the six types of distractions in the simulation experiment have a significant discrimination effect.The use of distraction warning can significantly improve the level of traffic safety,alleviate distracted driving behaviors and impact on road traffic safety.