Abstract: An innovation feature of this study is modeling real-time car-following risk based on probability computation.First of all，this paper divides the process of collision for car-following into two stages：leading vehicle decelerates and the following vehicle can not avoid the collision on the condition of it.On the basis of that，the probability of rear-end collision which is used to express the risk of car-following is calculated by the method of total probability.What’s more，a deceleration probability density function model is developed which is based on the data coming from video detector.The analysis of the condition on which collision occur while leading vehicle decelerating，brake system response time and sticking coefficient are taken into account and this make the model forceful to reflect the real traffic situation.At last，a case study is given by using Hang Zhou data.

Key words: traffic safety, car-following risk estimate, probability computation, video detect

摘要： 首次将追尾事故发生的过程分为前车减速与前车减速条件下的碰撞两起事件，运用条件概率的思想求出追尾事故发生的概率来表征跟车风险。首先从安全的本质出发提出了跟车风险的表述方法；再从实际道路采集减速度数据，根据实际数据建立密度函数模型。在分析前车减速条件下追尾碰撞条件时，从汽车地面力学理论的角度补充考虑了制动系作用时间、附着系数等影响因素，使模型更接近实际状况，并且将判断的标准由停车距离扩展到了制动全过程的位移。最后，运用视频检测方法提取杭州市道路交通运行数据对估计方法进行了试算。

关键词: 交通安全, 跟车风险估计, 概率计算, 视频检测