Institute of Transportation, MOTC

      Opposing left turn conflict is the scenario that both left turn vehicle and opposing straight vehicle have the legal righ t of way. T he reason for using this conflict as our research scenario, is according to the statistics from the ministry of co mmunications in Taiwan, side collision is the most common type of accident, which manifests the value of the safety evaluation of r elated conflict type. In the past decades, resea rchers usually evaluate the road safety by collision data. However, by the en hancement of computing power of computer, using the trajectory data, which is extracted from the video data with Artificial Intelli gence, has become value in the field of transpor tation safety recently. C ompared to the traditional safety analysis by collision data, it not only spends lower cost, but also get the stable results due to numerous data than original method in the constant period . Time to Collision (TTC) and Post Encroac hment(PET) are two of the best known i ndicators, and widely used in traffic conflict analysis. However, it might overestimated or underestimated the degree of hazard of conflict events, especially in the scen ario of opposing left turn conflict, by the comp lex real situation and the unreal pred iction method. To tackle with the problems, our research proposes t wo advanced indicators based on the different trajectory prediction methods, which take the changing of vehicle status and the road geometry facilities into account . As stated by research results, the advanced indicators are more adapted than TTC and PET, to better measure the hazard in specific type of opposing left turn conflict.