Institute of Transportation, MOTC

        Current highway bus operators have installed in-vehicle CCTVs to record bus driver’s behavior during transportation. It will be meaningful to enhance safety if these digital video data can be used to help bus drivers to avoid from abnormal driving behavior.

         Considering on-site hardware and computational resources this study has chosen Kamalan bus company as the use case for empirical analysis. Having discussed with top managers at Kamalan bus company, this study used R language to present a CNN-based model with continuous iterations and tests. Empirical results showed that the accurate rate reached 94% and achieved the goal of feasibility of recognizing bus driver’s behavior by using digital video data.

        With the help of visualization tool Power BI this study used rear-end operation data to map with bus driver’s behavior results. It is proven that Kamalan bus company awards recognition to this model because of positive responses of bus drivers in company. For example, overloaded bus drivers will be interviewed to understand why they performed abnormal driving behavior. Personal persuade measures or rescheduling work shit may be taken to improve driving behavior for transportation safety.