利用姿态信息实现异常行为检测
|
摘要
异常行为检测广泛应用于安防、智能交通、机场监视、监考等领域,但异常行为数据难以获取,算法准确率较低。为了应对上述问题,提出一个基于对抗自编码思想的两路异常检测网络。其中,一路子网络利用像素信息,关注行为发生的整体环境。另一路子网络则利用姿态信息,关注人体行为。然后对两个子网络的结果进行混合,得到异常行为检测的结果。最后,在CUHK Avenue和UCSD Ped数据集上验证结果。
Anomaly detection has been widely used in security, intelligent, invigilation, etc. But the abnormal data is difficult to obtain and the accuracies of the current algorithms are not high. To address these problems, proposes a new model, which consisting of two adversarial autoencoders-like(AAE-like) sub-networks. One sub-network focuses on the environment by processing appearance information. Another sub-network focuses on the human behavior by utilizing the human pose information in the frames. Then the results of the two sub-networks are combined to obtain the result. Finally, t evaluates the proposed model on CUHK Avenue and UCSD Ped datasets.
Anomaly detection has been widely used in security, intelligent, invigilation, etc. But the abnormal data is difficult to obtain and the accuracies of the current algorithms are not high. To address these problems, proposes a new model, which consisting of two adversarial autoencoders-like(AAE-like) sub-networks. One sub-network focuses on the environment by processing appearance information. Another sub-network focuses on the human behavior by utilizing the human pose information in the frames. Then the results of the two sub-networks are combined to obtain the result. Finally, t evaluates the proposed model on CUHK Avenue and UCSD Ped datasets.
引文
[1]杨锐，罗宾，郝叶林，常津津.一种基于深度学习的异常行为识别方法[J].五邑大学学报（自然科学版），2018:27-34.
[2]Akcay S,Atapour-Abarghouei A,P. Breckon T. GANomaly:Semi-Supervised Anomaly Detection via Adversarial Training[J]. ACCV,2018.
[3]Herath S,Harandi M,Porikli F. Going Deeper into Action Recognition:A Survey[J]. Image and Vision Computing,2017:4-21.
[4]Simonyan K,Zisserman A. Two-Stream Convolutional Networks for Action Recognition in Videos[J]. Advances in Neural Information Processing Systems,2014:568-576.
[5]Feichtenhofer C,Pinz A,Wildes R. P. Spatiotemporal Residual Networks for Video Action Recognition[J]. CVPR,2016.
[6]Feichtenhofer C,Pinz A,Wildes R. P. Spatiotemporal Multiplier Networks for Video Action Recognition[J]. CVPR,2017.
[7]Limin W,Yuanjun X,Zhe W,Yu Q,Dahua L,Xiaoou T,et al. Temporal Segment Networks:Towards Good Practices for Deep Action Recognition[J]. European Conference on Computer Vision,2016:20-36.
[8]Tran D,Bourdev L,Fergus R,Torresani L,Paluri M. Learning Spatiotemporal Features with 3D Convolutional Networks[J]. ICCV,2015.
[9]Varol G,Laptev I,Schmid C. Long-term Temporal Convolutions for Action Recognition[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence,2017:1-1.
[10]Xu H,Das A,Saenko K. R-C3D:Region Convolutional 3D Network for Temporal Activity Detection[J]. ICCV,2017.
[11]Wen L,Luo W,Lian D,Gao S. Future Frame Prediction for Anomaly Detection-A New Baseline[J]. IEEE Conference on Computer Vision and Pattern Recognition,2018.
[12]Kiran B,Thomas D,Ranjith P. An Overview of Deep Learning Based Methods for Unsupervised and Semi-Supervised Anomaly Detection in Videos[J]. Journal of Imaging,2018:36.
[13]王恬，李庆武，等.利用姿势估计实现人体异常行为识别[J].仪器仪表学报，2016:2366-2372.
[14]Morais1 R,Le V,Tran T,Saha B,Mansour M,Venkatesh S. Learning Regularity in Skeleton Trajectories for Anomaly Detection in Videos[Z]. unpublished.
[15]Sun K,Xiao B,Liu D,Jingdong W. Deep High-Resolution Representation Learning for Human Pose Estimation[Z]. unpublished.
[16]Insafutdinov E,Pishchulin L,Andres B,Andriluka M,Schiele B. DeeperCut:A Deeper,Stronger,and Faster Multi-Person Pose Estimation Model[J]. ECCV,2016.
[17]Hao-Shu F,Shuqin X,Yu-Wing T,Cewu L. RMPE:Regional Multi-person Pose Estimation[J]. ICCV,2017.
[18]Insafutdinov E,Andriluka M,Pishchulin L,Tang S,Levinkov E,Andres B,Schiele B. ArtTrack:Articulated Multi-Person Tracking in the Wild[J]. CVPR,2017.
[19]Cao Z,Simon T,Wei S E,Sheikh Y. Realtime Multi-Person 2D Pose Estimation Using Part Affinity Fields[J]. CVPR,2016.
[20]Goodfellow I J,Pouget-Abadie J,Mirza M,Xu B,Warde-Farley D,Ozair S,et al. Generative Adversarial Nets[Z]. MIT Press,2014.
[21]Makhzani A,Shlens J,Jaitly N,Goodfellow I,Frey B. Adversarial Autoencoders[J],2015.
[22]Ioffe S,Szegedy C. Batch Normalization:Accelerating Deep Network Training by Reducing Internal Covariate Shift[C]. International Conference on Machine Learning,2015.
[23]YuXin W,Kaiming H. Group Normalization[J],2018.
[24]Kim J,Grauman K. Observe Locally,Infer Globally:a Space-Time MRF for Detecting Abnormal Activities with Incremental Updates[J].CVPR,2009.
[2]Akcay S,Atapour-Abarghouei A,P. Breckon T. GANomaly:Semi-Supervised Anomaly Detection via Adversarial Training[J]. ACCV,2018.
[3]Herath S,Harandi M,Porikli F. Going Deeper into Action Recognition:A Survey[J]. Image and Vision Computing,2017:4-21.
[4]Simonyan K,Zisserman A. Two-Stream Convolutional Networks for Action Recognition in Videos[J]. Advances in Neural Information Processing Systems,2014:568-576.
[5]Feichtenhofer C,Pinz A,Wildes R. P. Spatiotemporal Residual Networks for Video Action Recognition[J]. CVPR,2016.
[6]Feichtenhofer C,Pinz A,Wildes R. P. Spatiotemporal Multiplier Networks for Video Action Recognition[J]. CVPR,2017.
[7]Limin W,Yuanjun X,Zhe W,Yu Q,Dahua L,Xiaoou T,et al. Temporal Segment Networks:Towards Good Practices for Deep Action Recognition[J]. European Conference on Computer Vision,2016:20-36.
[8]Tran D,Bourdev L,Fergus R,Torresani L,Paluri M. Learning Spatiotemporal Features with 3D Convolutional Networks[J]. ICCV,2015.
[9]Varol G,Laptev I,Schmid C. Long-term Temporal Convolutions for Action Recognition[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence,2017:1-1.
[10]Xu H,Das A,Saenko K. R-C3D:Region Convolutional 3D Network for Temporal Activity Detection[J]. ICCV,2017.
[11]Wen L,Luo W,Lian D,Gao S. Future Frame Prediction for Anomaly Detection-A New Baseline[J]. IEEE Conference on Computer Vision and Pattern Recognition,2018.
[12]Kiran B,Thomas D,Ranjith P. An Overview of Deep Learning Based Methods for Unsupervised and Semi-Supervised Anomaly Detection in Videos[J]. Journal of Imaging,2018:36.
[13]王恬，李庆武，等.利用姿势估计实现人体异常行为识别[J].仪器仪表学报，2016:2366-2372.
[14]Morais1 R,Le V,Tran T,Saha B,Mansour M,Venkatesh S. Learning Regularity in Skeleton Trajectories for Anomaly Detection in Videos[Z]. unpublished.
[15]Sun K,Xiao B,Liu D,Jingdong W. Deep High-Resolution Representation Learning for Human Pose Estimation[Z]. unpublished.
[16]Insafutdinov E,Pishchulin L,Andres B,Andriluka M,Schiele B. DeeperCut:A Deeper,Stronger,and Faster Multi-Person Pose Estimation Model[J]. ECCV,2016.
[17]Hao-Shu F,Shuqin X,Yu-Wing T,Cewu L. RMPE:Regional Multi-person Pose Estimation[J]. ICCV,2017.
[18]Insafutdinov E,Andriluka M,Pishchulin L,Tang S,Levinkov E,Andres B,Schiele B. ArtTrack:Articulated Multi-Person Tracking in the Wild[J]. CVPR,2017.
[19]Cao Z,Simon T,Wei S E,Sheikh Y. Realtime Multi-Person 2D Pose Estimation Using Part Affinity Fields[J]. CVPR,2016.
[20]Goodfellow I J,Pouget-Abadie J,Mirza M,Xu B,Warde-Farley D,Ozair S,et al. Generative Adversarial Nets[Z]. MIT Press,2014.
[21]Makhzani A,Shlens J,Jaitly N,Goodfellow I,Frey B. Adversarial Autoencoders[J],2015.
[22]Ioffe S,Szegedy C. Batch Normalization:Accelerating Deep Network Training by Reducing Internal Covariate Shift[C]. International Conference on Machine Learning,2015.
[23]YuXin W,Kaiming H. Group Normalization[J],2018.
[24]Kim J,Grauman K. Observe Locally,Infer Globally:a Space-Time MRF for Detecting Abnormal Activities with Incremental Updates[J].CVPR,2009.