Accident Detection Method Integrating Progressive Domain Adaptation and Cross-Attention

doi:10.3778/j.issn.1002-8331.2311-0109

Abstract

Abstract: Road traffic accidents are one of the major causes of property damage, casualties, and traffic congestion. Rapid and accurate detection of accidents is crucial for improving emergency response speed and reducing the fatality rate. In recent years, vision?based accident detection methods have become the mainstream, achieving significant progress through the application of deep learning techniques. However, existing methods suffer from low accuracy in domestic surveillance scenes due to significant domain shift between the open-source training data and domestic surveillance accident video data. To avoid the laborious task of constructing and annotating a large-scale domestic surveillance accident dataset, this paper proposes a domain adaptation dual-stream network that fully utilizes existing open-source accident datasets to achieve road accident detection in domestic surveillance scenes. Specifically, this paper develops a coarse-to-fine progressive domain adaptation training approach to guide the model in capturing domain-invariant accident appearance features. After that, a cross-attention fusion network is employed to establish correlations and interactions between appearance and motion features, alleviating feature discrepancies caused by scene variations. Experimental results demonstrate that the proposed method achieves high detection performance in domestic surveillance scenes, validating its effectiveness and reliability. This research is expected to provide important technical support for enhancing public safety and improving traffic efficiency.

Key words: intelligent transportation, computer vision, accident detection, domain adaptation dual-stream network

摘要： 道路交通事故是导致财产损失、人员伤亡和交通拥堵的主要原因之一。快速准确地检测事故对于提高应急救援速度和降低人员伤亡率至关重要。近年来，基于视觉的事故检测方法成为主流，且随深度学习技术的应用取得了显著进展。然而，现有方法应用于国内监控场景时存在精度不高的问题，主要原因是用于模型训练的开源数据与国内监控事故视频数据间存在较大的域偏移。为了避免重新构建并标注一个大型国内监控事故数据集的繁琐性，提出了一种域适应双流网络，充分利用现有开源事故数据集实现面向国内监控场景的道路事故检测。具体地提出了一种“由粗到细”递进式域适应方法，指导模型以捕捉域不变事故外观特征。构建交叉注意力融合网络实现外观特征和运动特征的关联与交互，缓解因场景差异导致的特征偏差问题。实验结果表明，所提方法在国内监控场景下取得了较高的检测性能，验证了其有效性和可靠性。有望为提升人民生命安全和交通效率提供了重要的技术支持。

关键词: 智能交通, 计算机视觉, 事故检测, 域适应双流网络

ZHANG Qi, ZHOU Wei, HU Weichao, YU Pengcheng. Accident Detection Method Integrating Progressive Domain Adaptation and Cross-Attention[J]. Computer Engineering and Applications, 2025, 61(6): 349-360.

张奇, 周威, 胡伟超, 于鹏程. 融合递进式域适应和交叉注意力的事故检测方法[J]. 计算机工程与应用, 2025, 61(6): 349-360.

References

[1] 李海涛, 李志慧, 王鑫, 等. 基于时间卷积自编码网络的实时交通事件自动检测方法[J]. 中国公路学报, 2022, 35(6): 265-276.
LI H T, LI Z H, WANG X, et al. Real-time automatic method of detecting traffic incidents based on temporal convolutional autoencoder network[J]. China Journal of Highway and Transport, 2022, 35(6): 265-276.
[2] 王晨, 周威, 严隽逸, 等. 一种用于道路交通事故自动检测的改进双流网络[J]. 中国公路学报, 2023, 36(5): 185-196.
WANG C, ZHOU W, YAN J Y, et al. Improved two-stream network for vision-based traffic accident detection[J]. China Journal of Highway and Transport, 2023, 36(5): 185-196.
[3] 王晨, 周威, 章世祥. 一种特征融合的视频事故快速检测方法[J]. 交通运输工程与信息学报, 2022, 20(1): 31-38.
WANG C, ZHOU W, ZHANG S X. A feature fusion deep learning framework for video-based crash detection systems[J]. Journal of Transportation Engineering and Information, 2022, 20(1): 31-38.
[4] 拜佩, 李金屏. 一种基于视频的交通事故检测方法[J]. 济南大学学报 (自然科学版), 2012, 26(3): 282-286.
BAI P, LI J P. A video-based method of traffic accident detection[J]. Journal of University of Jinan (Science and Technology), 2012, 26(3): 282-286.
[5] SADEK S, AL-HAMADI A, MICHAELIS B, et al. Real-time automatic traffic accident recognition using HFG[C]//Proceedings of the 20th International Conference on Pattern Recognition, 2010: 3348-3351.
[6] IJJINA E P, CHAND D, GUPT S, et al. Computer vision-based accident detection in traffic surveillance[C]//Proceedings of the International Conference on Computing, Communication and Networking Technologies, 2019: 1-6.
[7] MAALOUL B, TALEB-AHMED A, NIAR S, et al. Adaptive video-based algorithm for accident detection on highways[C]//Proceedings of the IEEE International Symposium on Industrial Embedded Systems, 2017: 1-6.
[8] ARCEDA V M, RIVEROS E L. Fast car crash detection in video[C]//Proceedings of the 2018 XLIV Latin American Computer Conference (CLEI), 2018: 632-637.
[9] 陈斌, 金炜东. 四类交通事故检测算法的分析[J]. 交通科技与经济, 2005(3): 50-52.
CHEN B, JIN W D. Analysis of four category accident detection algorithms[J]. Technology and Economy in Areas of Communication, 2005(3): 50-52.
[10] LU Z, ZHOU W, HANG S, et al. A new video-based crash detection method: balancing speed and accuracy using a feature fusion deep learning framework[J]. Journal of Advanced Transportation, 2020(8): 8848874.
[11] WANG C, DAI Y L, ZHOU W, et al. A vision-based video crash detection framework for mixed traffic flow environment considering low-visibility condition[J]. Journal of Advanced Transportation, 2020(1): 9194028.
[12] HUANG X, HE P, RANGARAJAN A, et al. Intelligent intersection: two-stream convolutional networks for real-time near accident detection in traffic video[J]. ACM Transactions on Spatial Algorithms and Systems, 2019, 6(2): 1-28.
[13] BATANINA E, BEKKOUCH I, YOUSSRY Y, et al. Domain adaptation for car accident detection in videos[C]//Proceedings of the 2019 Ninth International Conference on Image Processing Theory, Tools and Applications, 2019: 1-6.
[14] SINGH D, MOHAN C K. Deep spatio-temporal representation for detection of road accidents using stacked autoencoder[J]. IEEE Transactions on Intelligent Transportation Systems, 2018, 20(3): 879-887.
[15] LUO W, WEN L, GAO S. Remembering history with convolutional LSTM for anomaly detection[C]//Proceedings of the 2017 IEEE International Conference on Multimedia and Expo, 2017: 439-444.
[16] DOSHI K, YILMAZ Y. Online anomaly detection in surveillance videos with asymptotic bounds on false alarm rate[J]. Pattern Recognition, 2021, 114(2): 107865.
[17] SHAH A P, LAMARE J B, NGUYEN-ANH T, et al. CADP: a novel dataset for CCTV traffic camera based accident analysis[C]//Proceedings of the 15th IEEE International Conference on Advanced Video and Signal Based Surveillance, 2018: 1-9.
[18] WANG M, DENG W. Deep visual domain adaptation: a survey[J]. Neurocomputing, 2018, 312: 135-153.
[19] KI Y K, LEE D Y. A traffic accident recording and reporting model at intersections[J]. IEEE Transactions on Intelligent Transportation Systems, 2007, 8(2): 188-194.
[20] ZHONG J X, LI N, KONG W, et al. Graph convolutional label noise cleaner: train a plug-and-play action classifier for anomaly detection[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2019: 1237-1246.
[21] FENG J C, HONG F T, ZHENG W S. Mist: multiple instance self-training framework for video anomaly detection[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2021: 14009-14018.
[22] LUO W, LIU W, GAO S. Remembering history with convolutional LSTM for anomaly detection[C]//Proceedings of the 2017 IEEE International Conference on Multimedia and Expo (ICME), 2017: 439-444.
[23] LEE S, KIM H G, RO Y M. STAN: spatio-temporal adversarial networks for abnormal event detection[C]//Proceedings of the 2018 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), 2018: 1323-1327.
[24] GONG D, LIU L, LE V, et al. Memorizing normality to detect anomaly: memory-augmented deep autoencoder for unsupervised anomaly detection[C]//Proceedings of the IEEE/CVF International Conference on Computer Vision, 2019: 1705-1714.
[25] WANG X, CHE Z, JIANG B, et al. Robust unsupervised video anomaly detection by multipath frame prediction[J]. IEEE Transactions on Neural Networks and Learning Systems, 2021, 33(6): 2301-2312.
[26] DOSHI K, YILMAZ Y. Online anomaly detection in surveillance videos with asymptotic bound on false alarm rate[J]. Pattern Recognition, 2021, 114: 107865.
[27] ZHOU W, WEN L, ZHAN Y, et al. An appearance-motion network for vision-based crash detection: improving the accuracy in congested traffic[J]. IEEE Transactions on Intelligent Transportation Systems, 2023.
[28] SULTANI W, CHEN C, SHAH M. Real-world anomaly detection in surveillance videos[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2018: 6479-6488.
[29] LIU W, LUO W, LIAN D, et al. Future frame prediction for anomaly detection-a new baseline[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2018: 6536-6545.
[30] HE K, ZHANG X, REN S, et al. Deep residual learning for image recognition[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2016: 770-778.
[31] GANIN Y, LEMPITSKY V. Unsupervised domain adaptation by backpropagation[C]//Proceedings of the International Conference on Machine Learning, 2015: 1180-1189.
[32] KANG G, JIANG L, YANG Y, et al. Contrastive adaptation network for unsupervised domain adaptation[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2019: 4893-4902.
[33] 赵鹏飞, 李艳玲, 林民. 结合胶囊网络的领域适应意图识别[J]. 计算机工程与应用, 2021, 57(21): 188-194.
ZHAO P F, LI Y L, LIN M, et al. Intent detection of domain adaptation combined with capsule network[J]. Computer Engineering and Applications, 2021, 57(21): 188-194.
[34] 刘华玲, 皮常鹏, 赵晨宇，等. 基于深度域适应的跨域目标检测算法综述[J]. 计算机工程与应用, 2023, 59(8): 1-12.
LIU H L, PI C P, ZHAO C Y, et al. Review of cross-domain object detection algorithms based on depth domain adaptation[J]. Computer Engineering and Applications, 2023, 59(8): 1-12.
[35] 赵雪冰, 王俊杰. 基于改进DeeplabV3+和迁移学习的桥梁裂缝检测[J]. 计算机工程与应用, 2023, 59(5): 262-269.
ZHAO X B, WANG J J. Bridge crack detection based on improved DeeplabV3+ and migration learning[J]. Computer Engineering and Applications, 2023, 59(5): 262-269.
[36] CHEN C F R, FAN Q, PANDA R. CrossViT: cross-attention multi-scale vision transformer for image classification[C]//Proceedings of the IEEE/CVF International Conference on Computer Vision, 2021: 357-366.
[37] WANG L, XIONG Y, WANG Z, et al. Temporal segment networks for action recognition in videos[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2018, 41(11): 2740-2755.
[38] KIM H, LEE K, HWANG G, et al. Crash to not crash: learn to identify dangerous vehicles using a simulator[C]//Proceedings of the AAAI Conference on Artificial Intelligence, 2019: 978-985.
[39] LIN J, GAN C, HAN S. TSM: temporal shift module for efficient video understanding[C]//Proceedings of the IEEE/CVF International Conference on Computer Vision, 2019: 7083-7093.
[40] 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, 2017: 6299-6308.
[41] HAJRI F, FRADI H. Vision transformers for road accident detection from dashboard cameras[C]//Proceedings of the 2022 18th IEEE International Conference on Advanced Video and Signal Based Surveillance (AVSS), 2022: 1-8.
[42] KARIM M M, LI Y, QIN R, et al. A dynamic spatial-temporal attention network for early anticipation of traffic accidents[J]. IEEE Transactions on Intelligent Transportation Systems, 2022, 23(7): 9590-9600.
[43] HOWARD A, SANDLER M, CHU G, et al. Searching for mobilenetv3[C]//Proceedings of the IEEE/CVF International Conference on Computer Vision, 2019: 1314-1324.