基于无人机航拍轨迹数据的交通冲突严重程度指标对比分析  

Comparative Analysis of Surrogate Safety Measures Based on Drone Dataset of Vehicle Trajectories

作　　者：张晨晨 廖彦锟 ZHANG Chenchen;LIAO Yankun

机构地区：[1]上海建科工程咨询有限公司,上海200032 [2]同济大学道路与交通工程教育部重点实验室,上海201804

出　　处：《上海公路》2025年第1期173-181,M0009,共10页Shanghai Highways

基　　金：上海市国际科技合作基金项目(24170742100)。

摘　　要：囿于传统交通信息采集技术,既有的交通冲突严重程度指标(简称“冲突指标”)的适用场景、评价效果、交通流特征敏感性等特性缺乏基于大规模数据集的系统验证与对比分析。而基于无人机航拍视频提取的车辆轨迹数据则具有采样频率高、数据类型丰富、数据规模大等优势。基于指标类型及其常用性,选取TTC、PET、PICUD与DRAC四个冲突指标,明确了合理的冲突判别阈值,并利用基于无人机航拍视频构建的highD轨迹数据集,进行了冲突识别性能验证及对比分析。结果表明,PICUD具有最高的冲突识别敏感性,而DRAC最低;基于TTC识别出的冲突数量与交通流拥堵程度存在显著的正相关性。研究成果丰富了交通冲突技术理论体系的内涵,为基于交通冲突技术开展的交通安全分析研究中的冲突指标及其阈值的选取提供了依据。Constrained by traditional traffic information collection technology,the existing surrogate safety measures lack systematic verification and comparative analysis based on large-scale datasets in terms of application scenarios,evaluation effects,and sensitivity to certain features of traffic flow.The vehicle trajectory data extracted from drone aerial videos have advantages of high sampling rate,rich data types,and large data scale.Based on the types and their commonality,four surrogate safety measures,TTC,PET,PICUD,and DRAC,were selected,and reasonable conflict determination thresholds were clarified.The conflict identification performance verification and comparative analysis were conducted using the highD dataset constructed based on drone aerial videos.The results show that PICUD has the highest conflict identification sensitivity,while DRAC has the lowest;the number of conflicts identified based on TTC has a significant positive correlation with the degree of traffic congestion.This research enriches the theoretical framework of traffic conflict techniques and provided a basis for the selection of surrogate safety measures and their thresholds in traffic safety analysis based on traffic conflicts technique.

关 键 词：无人机 车辆轨迹 交通冲突技术 交通冲突严重程度指标 

分 类 号：TP3[自动化与计算机技术—计算机科学与技术]