基于车辆轨迹的信号交叉口机非冲突判别

选取长沙市中心区4个典型信号交叉口，利用视频轨迹追踪软件，提取右转机动车与直行非机动车的冲突交通流轨迹数据。以减速、换道等避险行为与可能发生碰撞(距离碰撞点时间小于2 s)为依据，采集机非冲突样本；选择距离碰撞最大时间(MTTC)和冲突时间差(TDTC)作为评价指标，提出一种基于交通流运行轨迹的改进型TTC(Time to Collision)法的机非冲突判别方法。选择某实例交叉口，分别采用TTC法、改进型TTC法和后侵入时间(PET)法，评估得到右转机动车与直行非机动车的冲突数，分别为7，24，22次。结果表明，传统的TTC法低估了交叉口的机非冲突水平，改进型TTC法的判别准确率提高了2.14倍，证明了方法的可行性。

Vehicle and Non-motorized Vehicle Traffic Conflict Recognition at Signalized Intersection Based on Vehicle Trajectory

This study investigated four typical signalized intersections in the downtown area of Changsha, and then used video trajectory tracking software to extract the conflicting trajectory data of right-turning vehicles and through movement non- motorized vehicles. Based on the risk- avoidance behaviors such as decelerating, lane changing and likely colliding (the time from the collision point is less than 2 seconds), the study collected 254 samples of vehicle and non-motorized vehicle traffic conflicts. To improve the traditional conflict recognition model, this study selected the maximum time to collision (MTTC) and the time difference to collision (TDTC) as the evaluation index, and proposed a modified TTC (Time To Collision) model to identify the vehicle and non- motorized vehicle conflict. A real intersection was then used for the empirical study. The number of traffic conflicts calculated by the TTC model, the modified TTC model and the post encroachment time (PET) model were 7, 24, and 22, respectively. The results indicate that the traditional TTC method underestimates the level of conflict between vehicles and non- motorized vehicles at intersections. The accuracy rate of improved TTC method is increased by 2.14 times, which proves the feasibility of the proposed method.