基于无人机视频的非机动车道交通冲突分析

doi:10.3778/j.issn.1002-8331.2206-0203

摘要/Abstract

摘要： 基于轨迹数据的交通冲突分析是道路交通流状态及安全风险评价的重要途径，针对当前电动自行车对混行交通流的影响研究不足的问题，提出基于无人机视频的目标时序轨迹提取方法和基于时序轨迹的行人与非机动车交通冲突识别与分析方法。以无人机监控视频作为输入，采用YOLOv5目标检测算法和DeepSORT多目标跟踪算法相结合，提取道路交叉口行人与非机动车的轨迹时序数据，经过坐标系转换，在每一时间步下，分析判断行人与非机动车间的正面冲突、追尾冲突和角度冲突。将冲突风险计算结果以热力图的形式进行可视化输出。实验结果表明，利用无人机视频提取的轨迹与实际轨迹的平均误差为7.48?cm，提出的研究方法具有较高的准确性；观测路段的交叉口处，角度冲突发生的频次最高且风险较大。研究成果能为道路交叉口数据采集提供新的思路，为路段安全风险评价提供有效的量化手段，为交通管理与控制提供决策支持。

关键词: 交通冲突, 无人机视频, 目标检测, 多目标跟踪, 非机动车

Abstract: Traffic conflict analysis based on trajectory data is an important way of evaluating road traffic flow status and safety risk. In view of the insufficient research on the impact of electric bicycles on mixed traffic flow, a method of extracting sequential trajectories of the target based on UAV video and a method of identifying and analyzing the traffic conflict between pedestrians and non-motor vehicles based on sequential trajectories are proposed. Taking UAV surveillance video as the input, the YOLOv5 target detection algorithm and DeepSORT multi-target tracking algorithm are combined to extract the sequential trajectories of pedestrians and non-motor vehicles at road intersections. At each time step, the head-on conflict, rear-end conflict and angle conflict between pedestrians and non-motor vehicles are analyzed and judged after the coordinate transformation. Then, the conflict risk calculation results are visually output in the form of a thermal diagram. The experimental results show that the average error between the trajectory data extracted by the UAV video and the actual trajectory is 7.48 cm, the proposed method has high accuracy. At the same time, the frequency of angle conflicts at the intersections of the observation section is the highest and the most severe. The research results can provide new ideas for road intersection data collection, effective quantitative means for road safety risk assessment as well as decision support for traffic management and control.

Key words: traffic conflict, UAV video, target detection, multi-target tracking, non-motor vehicle

周辉, 王维莉. 基于无人机视频的非机动车道交通冲突分析[J]. 计算机工程与应用, 2023, 59(16): 324-329.

ZHOU Hui, WANG Weili. Traffic Conflict Analysis of Non-Motor Vehicle Lanes Based on UAV Video[J]. Computer Engineering and Applications, 2023, 59(16): 324-329.

参考文献

[1] ISMAIL K，SC M，SAYED T，et al.Automated analysis of pedestrian-vehicle conflicts using video data[J].Transportation Research Record，2009，2140：44-54.
[2] SAUNIER N，SAYED T，ISMIAL K.Large-scale automated analysis of vehicle interactions and collisions[J].Transportation Research Record，2010，2147：42-50.
[3] SAYED T，ZAKI M H，AUTEY J.Automated safety diagnosis of vehicle-bicycle interactions using computer vision analysis[J].Safety Science，2013（59）：163-172.
[4] BEITEL D，STIPANCIC J，MANAUGH K，et al.Asessing safety of shared space using cyclist-pedestrian interactions and automated video conflict analysis[J].Transportation Research Part D：Transport and Environment，2018（65）：710-724.
[5] 文夏梅，傅立平，钟鸣，等.城市混合交通共享街道驾驶行为及行人过街安全研究[J].交通信息与安全，2020，38（5）：12-19.
WEN X M，FU L P，ZHONG M，et al.Analysis of driving behaviors and pedestrian crossing safety on urban roadway with mixed traffic flow[J].Journal of Transport Information and Safety，2020，38（5）：12-19.
[6] SUN H.Pedestrian movement tracking model in road environment based on UAV video[J].International Journal of Frontiers in Sociology，2020，2（7）：58-66.
[7] SUN X.Design of efficient human head statistics system in the large-angle overlooking scene[J].Electronics，2021，10（2）：45-49.
[8] 何宇，李丹.基于YOLOv5和DeepSort的实时地铁客流量检测[J].电子测试，2022，36（2）：46-48.
HE Y，LI D.Real-time subway ridership detection based on YOLOv5 and DeepSort[J].Electronic Test，2022，36（2）：46-48.
[9] 叶佳林，苏子毅，马浩炎，等.改进YOLOv3的非机动车检测与识别方法[J].计算机工程与应用，2021，57（1）：194-199.
YE J L，SU Z Y，MA H Y，et al.Improved YOLOv3 non-motor vehicle detection and recognition method[J].Computer Engineering and Applications，2021，57（1）：194-199.
[10] 李永上，马荣贵，张美月.改进YOLOv5s+DeepSORT的监控视频车流量统计[J].计算机工程与应用，2022，58（5）：271-279.
LI Y S，MA R G，ZHANG M Y.Traffic monitoring video vehicle volume statistics method based on improved YOLOv5s+DeepSORT[J].Computer Engineering and Applications，2022，58（5）：271-279.
[11] OPOWER H.Multiple viewgeometry in computer vision[J].Optics and Lasers in Engineering，2002，37（1）：85-86.
[12] BAI L，CHAN C Y，LIU P，et al.Identifying factors affecting the safety of mid-block bicycle lanes considering mixed 2-wheeled traffic flow[J].Traffic Injury Prevention，2017，18（7）：761-766.
[13] 陈思渊.基于无人机视频的交叉口行人运动跟踪及仿真建模[D].深圳：深圳大学，2019.
CHEN S Y.Tracking and modeling pedestrian movements and intersections based on UAV-video[D].Shenzhen：Shenzhen University，2019.
[14] LIU Y，STARZYK J A，ZHU Z.Optimized approximation algorithm in neural networks without overfitting[J].IEEE Transactions on Neural Networks，2008，19（6）：983-995.
[15] GUO Y Y，SAYED T，ZAKI M H.Exploring evasive action-based indicators for PTW conflicts in shared traffic facility environments[J].Transportation Engineering Part A Systems，2018，144（11）：04018065.
[16] 郑玉冰，马羊，程建川.基于轨迹数据的非机动车道内冲突事件自动识别与可视化[J].中国公路学报，2022，35（1）：71-84.
ZHENG Y B，MA Y，CHEN J C.Automated identification and visualization of conflict events in bike lanes using trajectory data[J].China Journal of Highway and Transport，2022，35（1）：71-84.