| Recent years,with the development of economic society,the ownership of vehicles is increasing everyday.Accompanied by huge amount of traffic accidents and potential traffic danger.A lot of traffic facilities was built for road safety,among these,road lighting at night is the most important one.Many studies have shown that 30%～40%of the traffic accidents were occurred at night,accident rate was higher than daytime.On the other hand,road lighting frequently focus on road segments,but ignore some risky sites such as roadside openings.Firstly,this paper used the questionnaire survey to research the subjective feeling of drivers when driving at night,searching the influence factors of night driving.The questionnaire included basic information survey,subjective feeling survey and behavior intention survey.The questionnaire was designed with Likert scale,and reliability and validity of samples were analyzed.After that,factor analysis was used to extract 7 common factors which influence night driving most.Fuzzy comprehensive evaluation was used for analyzing the influence of road lighting on drivers.Secondly,the UAV was used to collect data of vehicles passing roadside openings.Scatter diagrams of vehicles’ location and speed were drawn by the software MATLAB.Based on the data collected,by analyzing and comparing vehicle characteristics at daytime and night,this paper proposed a variable speed variation to represent the variation of vehicle speed.A road segment without road lighting was chosen as treatment group,and one of the three road segments was chosen to be control group.The speed variation was regarded as dependent variable,and lack of road lighting as the treatment.Then the DID model was applied to evaluate the treatment effect.Result indicated that lack of road lighting would lead to increasing of speed variation.At last,5 variables including illuminance,traffic volume,number of lanes,median and vehicle speed were regarded as input variables,speed variation was regarded as the output of the artificial neural network model.The BP algorithm was applied as solving method.The mean square error was used to evaluate the convergence of model.After training,sensibility of 5 input variables was analized.And relation curves of illuminance and speed variation under different situations were obtained.This paper also offered several proposals for road facility building. |
