To address the safety risk prevention requirements caused by right-of-way conflicts in the mixed traffic environment of motor and non-motorized vehicles at urban intersections, a collision risk prediction method for motor and non-motorized vehicles integrating trajectory clustering and deep learning was proposed in this paper to improve the accuracy of traffic safety assessment at intersections. First, multi-type vehicle trajectory data in drone videos were extracted based on the DataFromSky software to construct a high-precision spatiotemporal dataset containing vehicle coordinates, velocities, and accelerations, and high-risk scenario trajectories were screened through the feature analysis of risk scenarios. Second, the DBSCAN spatiotemporal trajectory clustering algorithm was adopted, and the spatial radius (REps) and minimum sample size (DMinPts) were optimized in combination with the silhouette coefficient method to divide the non-motorized vehicle trajectories into conservative and aggressive types. Then, an LSTM-based time series prediction model was designed to realize the multi-step prediction of future spatiotemporal trajectories of non-motorized vehicles, and a spatiotemporal collision risk quantification model based on Euclidean distance and TTC threshold was constructed to dynamically evaluate potential collision risks. The test results indicate that the DBSCAN clustering algorithm can effectively identify the driving behavior patterns of non-motorized vehicles (with a silhouette coefficient of 0.256 6), the LSTM model can accurately predict vehicle trajectories for the future three time steps, and the spatiotemporal collision risk quantification model can proactively identify high-risk interaction scenarios. The research results can provide a quantitative decision-making basis and effective technical support for real-time risk prevention and control in mixed traffic environments.
现有研究已在交通安全理论体系构建、事故数据统计分析等方面取得一定成果,但由于道路交叉口交通流运行复杂且动态变化,对行车风险的精准预测仍面临较大挑战。本文研究聚焦于城市交叉口的机非混合交通场景,通过基于密度的空间聚类算法(Density-Based Spatial Clustering of Applications with Noise,DBSCAN)区分非机动车驾驶行为(保守型/激进型),结合改进碰撞时间(Time-to-Collision,TTC)模型(融合方向角和加速度),可更加精准地量化斜向冲突风险;并应用长短时记忆神经网络(Long Short-Term Memory,LSTM)融合DBSCAN聚类生成的驾驶行为标签(保守型/激进型),将轨迹数据(速度、加速度)与驾驶行为特征联合输入,显著提升了激进型轨迹的预测敏感度。本文所提出的方法直接关联个体车辆的实时轨迹与行为模式,适用于复杂多变的交叉口环境,致力于交叉口行车安全提升实践[9-11],帮助交管部门及时发现高风险交叉口并采取干预措施,尤其在交通流量大、交叉口结构复杂、交通安全管理薄弱的区域,可及时为驾驶员提供碰撞风险预警,有效降低事故发生的概率,提升道路交通安全水平[12-13]。
National Bureau of Statistics. China statistical yearbook 2024[M]. Beijing: China Statistics Press, 2024.
[3]
ZengQ, WangQ F, ZhangK K, et al. Analysis of the injury severity of motor vehicle‑pedestrian crashes at urban intersections using spatiotemporal logistic regression models[J]. Accident Analysis & Prevention, 2023, 189: 107119.
[4]
BabaeiZ, Metin KuntM. A correlated random parameters ordered probit approach to analyze the injury severity of bicycle-motor vehicle collisions at intersections[J]. Accident Analysis & Prevention, 2024, 196: 107447.
HuLiwei, ZhangRuijie, ZhaoXueting, et al. Round about motor vehicle conflict risk identification model[J]. Journal of Transportation Systems Engineering and Information Technology, 2024, 24(1): 168-178.
WangYitong, ZhangAihong, BaoHantao, et al. Research on traffic safety at urban road intersections based on risk matrix[J]. Highways & Automotive Applications, 2023, 39(4): 28-34.
LiuMingjian, TanGuozhen, QianJingwei, et al. Research on traffic safety warning method based on space-time grid intersection[J]. Journal of Beijing University of Posts and Telecommunications, 2015, 38(2): 38-44.
WangBo, LiuChanghe, ZhangChi, et al. Crash risk early warning in highway work zone based on road surveillance camera[J]. Journal of Zhejiang University (Engineering Science), 2024, 58(6): 1221-1232.
[13]
杨路. 基于碰撞风险评估的智能车主动换道控制策略研究[D]. 长春: 吉林大学, 2024.
[14]
YangLu. Research on active lane change control strategy for intelligent vehicles based on collision risk assessment[D]. Changchun: Jilin University, 2024.
WangHailiang, ChengChong, GuanHong, et al. Study on length of deceleration lane of left off-ramp for urban underground interchanges based on traffic conflicts[J]. Journal of China & Foreign Highway, 2024, 44(1): 209-217.
WangHao, YanJie, GuoJianmin, et al. Research on multi-classification prediction of traffic accident severity on mountainous expressways[J]. Journal of China & Foreign Highway, 2025, 45(4): 203-210.
[23]
LanD T, YoonS. Trajectory clustering-based anomaly detection in indoor human movement[J]. Sensors, 2023, 23(6): 3318.
[24]
HuangZ H, GaoS B, CaiC X, et al. A rapid density method for taxi passengers hot spot recognition and visualization based on DBSCAN+[J]. Scientific Reports, 2021, 11: 9420.
WangZihao, FengJiao, LiPeng. Ship trajectory clustering based on MFDBSCAN algorithm combined with improved similarity[J]. Information Technology, 2025(1): 1-7, 14.
WuZhimin, HuangJue, XiangQi. Identification of freeway accident-prone sections based on K-means clustering algorithm[J]. Journal of China & Foreign Highway, 2022, 42(6): 260-264.
[31]
MohammadnazarA, ArvinR, KhattakA J. Classifying travelers' driving style using basic safety messages generated by connected vehicles: Application of unsupervised machine learning[J]. Transportation Research Part C: Emerging Technologies, 2021, 122: 102917.
[32]
WangJ Q, LiuK, LiH T. LSTM-based graph attention network for vehicle trajectory prediction[J]. Computer Networks, 2024, 248: 110477.
[33]
MinH T, XiongX Y, WangP Y, et al. A hierarchical LSTM-based vehicle trajectory prediction method considering interaction information[J]. Automotive Innovation, 2024, 7(1): 71-81.
[34]
ShiX J, ChenZ R, WangH, et al. Convolutional LSTM network: A machine learning approach for precipitation nowcasting[C]//Proceedings of the 29th Conference on Neural Information Processing Systems (NeurIPS), Montreal, Canada, 2015.
LiuZhi, LiXinyu, LiZhen, et al. Prediction method of surrounding rock deformation of highway tunnels based on Bayes-LSTM[J]. Journal of China & Foreign Highway, 2024, 44(1): 166-176.
JiXuewu, FeiCong, HeXiangkun, et al. Intention recognition and trajectory prediction for vehicles using LSTM network[J]. China Journal of Highway and Transport, 2019, 32(6): 34-42.
[39]
RossiL, AjmarA, PaolantiM, et al. Vehicle trajectory prediction and generation using LSTM models and GANs[J]. PLoS One, 2021, 16(7): e0253868.
[40]
LiR M, ZhongZ R, ChaiJ, et al. Autonomous vehicle trajectory combined prediction model based on CC-LSTM)[J]. International Journal of Fuzzy Systems, 2022, 24(8): 3798-3811.