[1]巢 新,吉根林,赵 斌,等.基于时空交互信息融合的车辆违规超车识别[J].南京师大学报(自然科学版),2026,49(02):85-97.[doi:10.3969/j.issn.1001-4616.2026.02.009]
　Chao Xin,Ji Genlin,Zhao Bin,et al.Spatiotemporal Interaction Information Fusion for Vehicle Illegal Overtaking Recognition[J].Journal of Nanjing Normal University(Natural Science Edition),2026,49(02):85-97.[doi:10.3969/j.issn.1001-4616.2026.02.009]
点击复制

基于时空交互信息融合的车辆违规超车识别()

《南京师大学报（自然科学版）》[ISSN:1001-4616/CN:32-1239/N]

卷:
49
期数:
2026年02期
页码:
85-97
栏目:
计算机科学与技术
出版日期:
2026-04-10

文章信息/Info

Title:
Spatiotemporal Interaction Information Fusion for Vehicle Illegal Overtaking Recognition
文章编号:
1001-4616(2026)02-0085-13
作者:
巢 新1; 2; 吉根林2; 赵 斌2; 麦丞程2; 王嘉琦2; 3
1.南京师范大学地理科学学院,江苏 南京 210023
2.南京师范大学计算机与电子信息学院/人工智能学院,江苏 南京 210023
3.南京师范大学 数学科学学院,江苏 南京 210023
Author(s):
Chao Xin1; 2; Ji Genlin2; Zhao Bin2; Mai Chengcheng2; Wang Jiaqi2; 3
1.School of Geography,Nanjing Normal University,Nanjing 210023,China
2.School of Computer and Electronic Information/School of Artificial Intelligence,Nanjing Normal University,Nanjing 210023,China
3.School of Mathematical Sciences,Nanjing Normal University,Nanjing 210023,China
关键词:
智能交通; 车辆违规超车行为识别; 时空交互信息融合; 多车辆交互建模; TimeSformer
Keywords:
intelligent transportation;  vehicle illegal overtaking recognition;  spatiotemporal interaction information fusion;  multi-vehicle interaction modeling;  TimeSformer
分类号:
U495
DOI:
10.3969/j.issn.1001-4616.2026.02.009
文献标志码:
A
摘要:
为提升车辆违规超车行为的识别精度,本文提出了基于时空交互信息融合的违规超车识别算法. 该算法以TimeSformer为主干模型,融合RGB图像、光流、深度图以及超车交互图四种模态信息,构建统一的超车信息图,从外观特征、运动特性、三维空间结构以及车辆间交互关系等多个维度对超车行为进行联合建模. 通过引入分离时空注意力机制以及多模态特征融合策略,有效刻画超车过程中目标车辆的动态演化特征及其与周围车辆之间的时空交互模式,从而弥补复杂交通场景下多车辆交互关系表述不足的问题. 在PREVENTION数据集上的实验结果表明,所提算法在违规超车识别任务中取得了94.04%的识别准确率,较多种主流基准算法表现出更优的识别性能,验证了多模态时空交互信息融合策略在复杂交通行为识别中的有效性.
Abstract:
To improve the accuracy of vehicle illegal overtaking recognition, this paper proposes spatiotemporal interaction information fusion for vehicle illegal overtaking recognition algorithm. The algorithm is built upon the TimeSformer architecture as the backbone model. Four types of modality information, namely RGB images, optical flow, depth maps, and overtaking interaction graphs, are integrated to construct a unified overtaking information graph. From multiple perspectives, including appearance features, motion information, 3D spatial structure, and inter-vehicle interaction relationships, the method performs joint modeling of overtaking behaviors. By introducing divided space-time attention mechanism and multi-modal feature fusion strategy, the proposed approach effectively captures the dynamic evolution of the target vehicle during the overtaking process as well as its spatiotemporal interactions with surrounding vehicles, thereby alleviating the insufficient representation of multi-vehicle interactions in complex traffic scenarios. Experimental results on the PREVENTION dataset show that the proposed algorithm achieves a recognition accuracy of 94.04% for illegal overtaking behaviors, outperforming several existing mainstream algorithms and validating the effectiveness of multimodal spatiotemporal interaction information fusion for complex traffic behavior recognition.

参考文献/References:

[1]Hong S,Yue T,Liu H. Vehicle energy system active defense:A health assessment of lithium-ion batteries[J]. International Journal of Intelligent System,2022; 37:10081-10099.
[2]Nalcakan Y,Bastanlar Y. Monocular vision-based prediction of cut-in manoeuvres with LSTM networks[C]//Proceedings of Science,Engineering Management and Information Technology. Ankara,Turkey:Springer,2022:111-123.
[3]Nalcakan Y,Bastanlar Y. Cut-in maneuver detection with self-supervised contrastive video representation learning[J]. Signal,Image and Video Processing,2023,17:2915-2923.
[4]Li Z,Wang Y,Zuo Z. Interaction-aware prediction for cut-in trajectories with limited observable neighboring vehicles[J]. IEEE Transactions on Intelligent Vehicles,2023,8(3):2148-2161.
[5]Ong C S,Connie T,Goh M K O. Vehicle overtaking detection using computer vision techniques[C]//Proceedings of International Symposium on Intelligent Robotics and Systems. Changsha,China:IEEE,2024.
[6]Marcomini K D,Brito V d C,Balestra G d C,et al. A novel approach to road safety:detecting illegal overtaking using smartphone cameras and deep learning for vehicle auditing[J]. Journal of Sensor and Actuator Networks,2025,14(1):10.
[7]Zhang T,Zou Y,Zhang X,et al. Data-driven based cruise control of connected and automated vehicles under cyber-physical system framework[J]. IEEE Transactions on Intelligent Transportation Systems,2021,22(10):6307-6319.
[8]Lyu N,Wen J,Duan Z,et al. Vehicle trajectory prediction and cut-in collision warning model in a connected vehicle environment[J]. IEEE Transactions on Intelligent Transportation Systems,2022,23(2):966-981.
[9]Lu Y,Huang L,Yao J,et al. Intention prediction-based control for vehicle platoon to handle driver cut-in[J]. IEEE Transactions on Intelligent Transportation Systems,2023,24(5):5489-5501.
[10]马永杰,程时升,马芸婷,等. 卷积神经网络及其在智能交通系统中的应用综述[J]. 交通运输工程学报,2021,21(4):48-71.
[11]杨建喜,郁超顺,李韧,等. 基于多周期组件时空神经网络的路网通行速度预测[J]. 交通运输系统工程与信息,2021,21(3):112-119+139.
[12]齐浩轩,曹弋,赵斌. 基于强化邻接图卷积的交叉口车辆轨迹预测模型[J]. 地球信息科学学报,2025,27(3):623-635.
[13]Rasch A,Flannagan C,Dozza M. When is it safe to complete an overtaking maneuver?modeling drivers' decision to return after passing a cyclist[J]. IEEE Transactions on Intelligent Transportation Systems,2024,25(11):15587-15599.
[14]Lu Y,Wang B,Huang L,et al. Modeling of driver cut-in behavior towards a platoon[J]. IEEE Transactions on Intelligent Transportation Systems,2022,23(12):24636-24648.
[15]Athree M,Jayasiri A. Vision-based automatic warning system to prevent dangerous and illegal vehicle overtaking[C]//Proceedings of International Research Conference on Smart Computing and Systems Engineering. Colombo,Sri Lanka:IEEE,2020.
[16]Wang P,Wu X,He X. Vibration-theoretic approach to vulnerability analysis of nonlinear vehicle platoons[J]. IEEE Transactions on Intelligent Transportation Systems,2023,24(10):11334-11344.
[17]Tran D,Bourdev L,Fergus R,et al. Learning spatio-temporal features with 3d convolutional networks[C]//Proceedings of the IEEE Conference on International Conference on Computer Vision. Santiago,Chile:IEEE,2015:4489-4497.
[18]Carreira J,Zisserman A. Quo vadis,action recognition?A new model and the kinetics dataset[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. Honolulu,Hawaii,USA:IEEE,2017:6299-6308.
[19]Wang L,Xiong Y,Wang Z,et al. Towards good practices for very deep two-stream convNets[PP]. 2015,arXiv:1507.02159.
[20]Chao X,Qi X,Ding R,et al. Vehicle lane change behavior recognition based on multi-scale three-stream 3D ResNets[J]. Multimedia Systems,2025,31:129.
[21]Sun D,Yang X,Liu M,et al. PWC-Net:CNNs for optical flow using pyramid,warping,and cost volume[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. Salt Lake City,Utah,USA:IEEE,2018.
[22]Yang L,Kang B,Huang Z,et al. Depth anything V2[C]//Proceedings of Neural Information Processing Systems. Vancouver,Canada,2024:21875-21911.
[23]Pandey S,Chen K,Dam E B. Comprehensive multimodal segmentation in medical imaging:combining yolov8 with sam and hq-sam models[C]//Proceedings of the IEEE International Conference on Computer Vision Workshops. Paris,France:IEEE,2023:2592-2598.
[24]Bertasius G,Wang H,Torresani L. Is space-time attention all you need for video understanding?[C]//Proceedings of the International Conference on Machine Learning. Vienna,Austria:PMLR,2021.
[25]Izquierdo R,Quintanar A,Parra I,et al. The prevention dataset:a novel benchmark for prediction of vehicles intentions[C]//Proceedings of the IEEE Intelligent Transportation Systems Conference. Auckland,New Zealand:IEEE,2019.
[26]Arnab A,Dehghani M,Heigold G,et al. ViViT:a video vision transformer[C]//Proceedings of the IEEE/CVF International Conference on Computer Vision. Montreal,Quebec,Canada:IEEE,2021:6836-6846.

备注/Memo

备注/Memo:
收稿日期:2025-11-19.
基金项目:国家自然科学基金项目(41971343)、江苏省前沿技术研发计划项目(BF2024005).
通讯作者:吉根林,博士,教授,博士生导师,研究方向:大数据挖掘与人工智能. E-mail:glji@njnu.edu.cn
更新日期/Last Update: 2026-04-10