论文部分内容阅读
Risky driving behavior of taxi drivers typically evaluated for full operation or sometimes sorted into occupied and empty running trips.In this thesis,we simultaneously analyze aggressive driving and distracted driving of taxi drivers under three different trip categories.Trip origin is considered a transition from without ride-order to with ride-order travelling or from with ride-order to occupied travelling,and a destination as a transition from occupied to without ride-order travelling and vice versa.Distracted driving is characterized by driver interference,driver mobile use and some entertainment aspects,while specific harmful and risky actions are considered for aggressive driving.High-resolution and real-time kinematic parameters of taxis were recorded by the invehicle recorder VBOX for overall 562 trips.The distracted driving parameters and aggressive driving actions were monitored through python data collector web application that was specially programmed for this particular research.Besides dual dash cam（i.e.,front and inside car camera）,drivers’ whole day driving history from their Chinese ride-hailing Di Di smart application was used to differentiate occupied trips,unoccupied trips with ride-order and unoccupied trips without rideorder.Structural Equation Modeling（SEM）is practiced in this paper to understand the influence of distracted driving indicators on aggressive driving behaviors.The multi-group model analysis of SEM indicated that handling distracted risky driving could control aggressive driving behavior up to 96% and 98% in unoccupied without ride-order trips and unoccupied trips with ride-order respectively.Moreover,Support vector machine（SVM）was trained on physiological signals（i.e.,facial expression data）for the recognition of aggressive driving behavior.The SVM model has examined the recognition ability of emotion data and facial micro-expression（ME）units.