基于轨迹数据的高速公路分流区冲突替代指标研究  

Development of Alternative Indicator of Conflict at Freeway Diverging Area Using Trajectory Data

作　　者：姜雪娇[1] 郭唐仪[1] 

机构地区：[1]南京理工大学自动化学院,南京210094

出　　处：《交通信息与安全》2016年第2期39-47,54,共10页Journal of Transport Information and Safety

基　　金：国家自然科学基金青年基金资助项目(51208261);教育部人文社会科学青年基金资助项目(12YCZH062);中国博士后基金(2015M572728)资助

摘　　要：非事故统计交通安全评价指标冲突时间(Time to Conflict,TTC)和后侵入时间(Post Encroachment Time,PET)在2车间距较近或接近冲突点车速较快时,会出现冲突严重程度判别错误。为提高冲突判别的准确性,以高速公路分流区为研究对象,提出基于行车轨迹的冲突严重程度判别方法。采用视频检测技术,对固定背景下的运动目标进行轨迹提取,利用改进的神经网络算法进行实时轨迹预测,根据预测轨迹坐标进行分流区冲突判别,并引入量化指标J以碰撞概率分析分流区2运动单元冲突严重程度。实例分析结果显示,量化指标J考虑避险行为对潜在冲突点出现时间的影响,准确性更高;以J指标预测所得两车最大碰撞概率为59.93%,比TTC下必定碰撞(100%),更接近于实际未碰撞观测情况。Using traditional off-incident traffic safety indicators such as Time to Conflict（TTC）and Post Encroachment Time（PET）may lead to a significant misjudgment of the severity of conflict if the two conflicting vehicles approach each other with a high speed or within a small distance.To improve the accuracy of conflict judgment,this paper proposes a trajectory-based method to evaluate the conflict severity at diverging areas of freeways.On the basis of vehicle trajectories extracted using a video detection technique from a fixed background,the driving trajectories are predicted through an improved neural network algorithm.Then,the conflicts at diverging areas of freeways are identified based on the predicted trajectories.Using the probability theory,the quantitative indicator Jis introduced to analyze the conflict severity.The results of a case study show that the identification accuracy of conflict severity is improved by using the proposed indicator J,which takes the danger-averting actions into consideration.The highest probability of the predicted conflict is59.93% based on J when compared to 100% based on TTC.The J-based probability is much closer to the observed traffic conflicts at freeway diverging areas.

关 键 词：交通安全 高速公路 分流区冲突 轨迹预测 安全替代指标 

分 类 号：U495[交通运输工程—交通运输规划与管理]