元学习和时空关系引导的视频异常检测方法

doi:10.3778/j.issn.1002-8331.2409-0408

摘要/Abstract

摘要： 随着监控设备的广泛应用和人们对公共安全的日益重视，视频异常检测已成为计算机视觉领域的研究热点。自编码器在解码过程中只关注预测帧的视觉特征与输入视频帧的对应关系，这使得模型对某些异常对象不敏感。同时，注意时空信息有助于更好地理解视频。提出了一种基于元学习的视频异常检测方法。在自编码器中引入元学习模块，利用其学习特性对特征进行编码、存储和更新，从而增强模型的泛化能力。设计了一个关注对象间时空关系的关注模块，并将其应用到未来帧预测网络中，以更准确地预测下一帧并检测异常事件。通过广泛的实验验证，该方法在UCSD Ped2、CUHK Avenue和ShanghaiTech视频异常检测数据集上取得了显著的性能提升，准确率分别达到了97.63%、88.41%和75.43%，显示了良好的准确性和鲁棒性。

关键词: 视频异常检测, 元学习, 时空注意机制, 未来帧异常检测

Abstract: With the wide application of monitoring equipment and people’s increasing attention to public safety, video anomaly detection has become a research hotspot in the field of computer vision. In the decoding process, the autoencoder only pays attention to the correspondence between the visual features of the predicted frame and the input video frame, which makes the model insensitive to some abnormal objects. At the same time, paying attention to temporal and spatial information helps to better understand the video. This paper proposes a video anomaly detection method based on meta-learning. Firstly, a meta-learning module is introduced into the autoencoder to encode, store and update the features with its learning characteristics, thus enhancing the generalization ability of the model. Secondly, this paper designs an attention module that focuses on the spatio-temporal relationship between objects and applies it to the future frame prediction network to predict the next frame more accurately and detects abnormal events. Finally, through extensive experimental verification, the proposed method has achieved significant performance improvement in UCSD Pedestrian 2, CUHK Avenue and ShanghaiTech video anomaly detection dataset, with accuracy rates reaching 97.63%, 88.41% and 75.43%, respectively. It shows good accuracy and robustness.

Key words: video anomaly detection, meta-learning, spatio-temporal attention mechanism, future frame anomaly detection

杜宇龙, 李洪均. 元学习和时空关系引导的视频异常检测方法[J]. 计算机工程与应用, 2025, 61(24): 197-205.

DU Yulong, LI Hongjun. Video Anomaly Detection Method Guided by Meta-Learning and Spatio-Temporal Relationship[J]. Computer Engineering and Applications, 2025, 61(24): 197-205.

参考文献

[1] 陈明一, 李洪均. 拓扑信息引导的视频异常行为检测方法[J]. 计算机工程与应用, 2024, 60(16): 228-235.
CHEN M Y, LI H J. Topology information guided video abn-ormal behavior detection method[J]. Computer Engineering and Applications, 2024, 60(16): 228-235.
[2] YU M, QUAN T Q, PENG Q L, et al. A model-based collaborate filtering algorithm based on stacked AutoEncoder[J]. Neural Computing and Applications, 2022, 34(4): 2503-2511.
[3] MASSOLI F V, FALCHI F, KANTARCI A, et al. MOCCA: multilayer one-class classification for anomaly detection[J]. IEEE Transactions on Neural Networks and Learning Systems, 2022, 33(6): 2313-2323.
[4] LU Y W, YU F, REDDY M K K, et al. Few-shot scene-adaptive anomaly detection[C]//Proceedings of the 16th Eur-opean Conference on Computer Vision. Cham: Springer, 2020: 125-141.
[5] FINN C, ABBEEL P, LEVINE S. Model-agnostic meta-learning for fast adaptation of deep networks[C]//Proceedings of the International Conference on Machine Learning, 2017.
[6] LV H, CHEN C, CUI Z, et al. Learning normal dynamics in videos with meta prototype network[C]//Proceedings of the 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2021: 15420-15429.
[7] ZHANG S Y, SONG X H, WANG C D, et al. Object-meta and MSGAE-MP: multi-dimensional video anomaly detection[J]. IET Computer Vision, 2025, 19(1): e12156.
[8] LIU W, LUO W X, LIAN D Z, et al. Future frame prediction for anomaly detection-a new baseline[C]//Proceedings of the 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2018: 6536-6545.
[9] GONG D, LIU L Q, LE V, et al. Memorizing normality to detect anomaly: memory-augmented deep autoencoder for unsupervised anomaly detection[C]//Proceedings of the 2019 IEEE/CVF International Conference on Computer Vision. Piscataway: IEEE, 2020: 1705-1714.
[10] 朱新瑞, 钱小燕, 施俞洲, 等. 长短期时间序列关联的视频异常事件检测[J]. 中国图象图形学报, 2024, 29(7): 1998-2010.
ZHU X R, QIAN X Y, SHI Y Z, et al. Video anomaly dete-ction with long-and-short-term time series correlations[J]. Journal of Image and Graphics, 2024, 29(7): 1998-2010.
[11] YANG Z W, LIU J, WU Z Y, et al. Video event restoration based on keyframes for video anomaly detection[C]//Proceedings of the 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2023: 14592-14601.
[12] 刘禹含, 吉根林, 张红苹. 基于骨架图与混合注意力的视频行人异常检测方法[J]. 计算机应用, 2024, 44(8): 2551-2557.
LIU Y H, JI G L, ZHANG H P. Video pedestrian anomaly detection method based on skeleton graph and mixed att-ention[J]. Journal of Computer Applications, 2024, 44(8): 2551-2557.
[13] 张红民, 颜鼎鼎, 田钱前. 改进时空图卷积网络的视频异常检测方法[J]. 光电工程, 2024, 51(5): 48-60.
ZHANG H M, YAN D D, TIAN Q Q. Improved spatio-temporal graph convolutional networks for video anomaly detection[J]. Opto-Electronic Engineering, 2024, 51(5): 48-60.
[14] 李凡长, 刘洋, 吴鹏翔, 等. 元学习研究综述[J]. 计算机学报, 2021, 44(2): 422-446.
LI F Z, LIU Y, WU P X, et al. A survey on recent advances in meta-learning[J]. Chinese Journal of Computers, 2021, 44(2): 422-446.
[15] WANG H C, ZHANG X D, HU Y T, et al. Few-shot semantic segmentation with democratic attention networks[C]//Proceedings of the 16th European Conference on Computer Vision. Cham: Springer, 2020: 730-746.
[16] PARK H, NOH J, HAM B. Learning memory-guided normality for anomaly detection[C]//Proceedings of the 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2020: 14360-14369.
[17] LE V T, KIM Y G. Attention-based residual autoencoder for video anomaly detection[J]. Applied Intelligence, 2023, 53(3): 3240-3254.
[18] CHO M, KIM T, KIM W J, et al. Unsupervised video anomaly detection via normalizing flows with implicit latent features[J]. Pattern Recognition, 2022, 129: 108703.
[19] HYUN W, NAM W J, LEE S W. Dissimilate-and-assimilate strategy for video anomaly detection and localization[J]. Neurocomputing, 2023, 522: 203-213.
[20] AICH A, PENG K C, ROY-CHOWDHURY A K. Cross-domain video anomaly detection without target domain adaptation[C]//Proceedings of the 2023 IEEE/CVF Winter Conference on Applications of Computer Vision. Piscataway: IEEE, 2023: 2578-2590.
[21] SINGH R, SAINI K, SETHI A, et al. STemGAN: spatio-temporal generative adversarial network for video anomaly detection[J]. Applied Intelligence, 2023, 53(23): 28133-28152.
[22] LI S F, CHENG Y, ZHAO L Y, et al. Anomaly detection with multi-scale pyramid grid templates[J]. Multimedia Tools and Applications, 2024, 83(4): 9929-9947.
[23] LI D H, NIE X S, GONG R, et al. Multi-branch GAN-based abnormal events detection via context learning in surveillance videos[J]. IEEE Transactions on Circuits and Systems for Video Technology, 2024, 34(5): 3439-3450.
[24] YE Q, SONG Z H, ZHAO Y Q, et al. Dual-channel autoencoder with key region feature enhancement for video anomalous event detection[J]. Neural Processing Letters, 2024, 56(3): 186.
[25] LI N J, CHANG F L, LIU C S. Spatial-temporal cascade autoencoder for video anomaly detection in crowded scenes[J]. IEEE Transactions on Multimedia, 2021, 23: 203-215.
[26] 潘理虎, 彭守信, 张睿, 等. 面向运动前景区域的视频异常检测[J]. 计算机应用, 2025, 45(4): 1300-1309.
PAN L H, PENG S X, ZHANG R, et al. Video anomaly detection for moving foreground regions[J]. Journal of Computer Applications, 2025, 45(4): 1300-1309.