# 基于改进的动态聚类的视频异常事件检测Video Abnormal Event Detection Based on Improved Dynamic Clustering

DOI: 10.12677/AIRR.2018.72009, PDF, HTML, XML, 下载: 632  浏览: 1,616  科研立项经费支持

Abstract: Abnormal event detection is an important part of intelligent surveillance systems, especially for complex surveillance video scenes. In recent years, many algorithms have been proposed to detect abnormal events. However, most of them need to set a series of parameters in the model during the modeling process, which is not only troublesome in arranging the parameters, but also the parameters need to be reset when changing the detect scene. This paper proposes an abnormal detection algorithm based on non-parametric models, constructs and maintains a vector set based on the motion trend vector merging method, and uses clustering to generate different event clusters, and proposes a pre-detection step to improve the detection effect of the algorithm in sparse scenes. Finally, some existing detection algorithms are selected for comparison experiments. The results show that the model proposed in this paper has certain advantages in detection rate and time performance.

1. 引言

· 对原算法在某些特定场景下检测出现偏差的原因进行了实验分析；

· 基于对问题的实验分析，分别采用基于运动趋势合并的方法来维护大小固定的字典集合以及预检测的方法来排除背景特征对检测的干扰。

Figure 1. Detection result in corresponding scenes

2. 相关工作

3. 问题分析

3.1. 同向异速漏检

${v}_{i}=\stackrel{¯}{M}×{\sum }_{j\in \text{grid}}g\left({D}_{j}\right),\text{\hspace{0.17em}}i=1,2,\cdots ,n$ (1)

$\left\{\begin{array}{l}\stackrel{¯}{M}=\frac{1}{nmt}{\sum }_{j\in \text{grid}}\sqrt{{v}_{{x}_{j}}^{2}+{v}_{{y}_{j}}^{2}}\\ {D}_{j}=\mathrm{arctan}\left(\frac{{v}_{{y}_{j}}}{{v}_{{x}_{j}}}\right)\\ g\left({D}_{j}\right)=\left\{\begin{array}{l}1,\text{\hspace{0.17em}}\text{\hspace{0.17em}}{D}_{j}\in \left[\frac{2\text{π}i}{n},\frac{2\text{π}i}{n}+1\right)\\ 0,\text{\hspace{0.17em}}\text{\hspace{0.17em}}\text{else}\end{array}\end{array}$ (2)

$\left\{\begin{array}{l}v=\left(1-\alpha \right)×{v}_{a}+\alpha ×{v}_{b}\\ r\left(v\right)=r\left({v}_{a}\right)+r\left({v}_{b}\right)\end{array}$ (3)

3.2. 稀疏场景误检

$\left\{\begin{array}{l}r\left({C}_{i}\right)={\sum }_{j\in {C}_{i}}r\left({v}_{j}\right)\\ w\left({C}_{i}\right)=\frac{r\left({C}_{i}\right)}{{\sum }_{i=1}^{K}{\sum }_{j\in {C}_{i}}\text{ }r\left({v}_{j}\right)}\end{array}$ (4)

4. 模型改进

4.1. 预处理及特征提取

${\sum }_{j\in \text{grid}}count\left(\sqrt{{v}_{{x}_{j}}^{2}+{v}_{{y}_{j}}^{2}}\right)>{c}_{th}$ (5)

$count\left(\sqrt{{v}_{{x}_{j}}^{2}+{v}_{{y}_{j}}^{2}}\right)=\left\{\begin{array}{l}1,\text{\hspace{0.17em}}\text{\hspace{0.17em}}\sqrt{{v}_{{x}_{j}}^{2}+{v}_{{y}_{j}}^{2}}>{M}_{th}\\ 0,\text{\hspace{0.17em}}\text{\hspace{0.17em}}\sqrt{{v}_{{x}_{j}}^{2}+{v}_{{y}_{j}}^{2}}<{M}_{th}\end{array}$ (6)

4.2. 集合维护及聚类

$v=f\left({v}_{a},{v}_{b}\right)$

$f\left({v}_{a},{v}_{b}\right)={\sum }_{i=1}^{n}\left(1-\alpha \right){v}_{{a}_{i}}+sign\left(i\right)\alpha {v}_{{b}_{i}}$ (7)

$sign\left(i\right)=\left\{\begin{array}{l}1,\text{\hspace{0.17em}}\text{\hspace{0.17em}}\text{\hspace{0.17em}}\text{\hspace{0.17em}}{v}_{{a}_{i}}<{v}_{{b}_{i}}\\ -1,\text{\hspace{0.17em}}\text{\hspace{0.17em}}{v}_{{a}_{i}}>{v}_{{b}_{i}}\end{array}$ (8)

${d}_{i}={\mathrm{max}}_{j\in {C}_{i}}dis\left({c}_{i},{v}_{j}\right)$ (9)

4.3. 异常检测

${\forall }_{i\in K}dis\left(v,{c}_{i}\right)>{d}_{i}$ (10)

${\exists }_{i\in K}\text{\hspace{0.17em}}s.t.\text{\hspace{0.17em}}\left\{\begin{array}{l}dis\left(v,{c}_{i}\right)<{d}_{i}\\ w\left({C}_{i}\right)<{w}_{th}\end{array}$ (11)

5. 实验

5.1. UCSD Ped2数据集

UCSD Ped2 dataset [28] 是一个被广泛认可及应用的视频异常检测数据集。其包含的图像序列均为背景相同且固定的监控视频场景，定义行人的运动为一般性运动，均是平行于摄像机的水平运动；而数据集定义的异常事件则包括自行车、汽车、滑板等与行人运动速度不在同一水平上的运动特征；数据集中包含有16个训练图像序列以及12个测试图像序列，图像大小均为360 × 240像素，每个grid中包含45 × 30 × 8个像素点。

UCSD数据集已有现有的评价体系，本文采用帧级别的异常检测来评价算法的检测效果，对于所有的视频帧将其视为01标签的单幅图像，若图像中存在异常事件且算法报警，则为检测正确，反之为误检。总的检测结果如表1所示。

Figure 2. Detection result of UCSD Ped2 dataset

Table 1. Detection comparison on UCSD Ped2 dataset

5.2. Subway数据集

Figure 3. Detection result of subway dataset

Table 2. Detection comparison on subway dataset

6. 总结

NOTES

*通讯作者。

 [1] Yogameena, B. and Priya, K.S. (2015) Synoptic Video Based Human Crowd Behavior Analysis for Forensic Video Surveillance. Eighth International Conference on Advances in Pattern Recognition, Kolkata, 4-7 January 2015, 1-6. https://doi.org/10.1109/ICAPR.2015.7050662 [2] Wang, L. and Dong, M. (2012) Real-Time Detection of Abnormal Crowd Behavior Using a Matrix Approximation-Based Approach. IEEE International Conference on Image Processing, Orlando, 30 September-3 October 2012, 2701-2704. [3] Yuan, Y., Fang, J. and Wang, Q. (2015) Online Anomaly Detection in Crowd Scenes via Structure Analysis. IEEE Transactions on Cybernetics, 45, 562-575. https://doi.org/10.1109/TCYB.2014.2330853 [4] Feng, J., Zhang, C. and Hao, P. (2012) Online Anomaly Detection in Videos by Clustering Dynamic Exemplars. International Conference on Image Processing, Orlando, 30 September-3 October 2012, 3097-3100. https://doi.org/10.1109/ICIP.2012.6467555 [5] Wang, H., Kläser, A. and Schmid, C. (2013) Dense Trajectories and Mo-tion Boundary Descriptors for Action Recognition. International Journal of Computer Vision, 103, 60-79. https://doi.org/10.1007/s11263-012-0594-8 [6] Wang, H. and Schmid, C. (2013) Action Recognition with Improved Trajectories. IEEE International Conference on Computer Vision, Sydney, December 2013, 3551-3558. https://doi.org/10.1109/ICCV.2013.441 [7] Wang, L., Qiao, Y. and Tang, X. (2015) Action Recognition with Trajecto-ry-Pooled Deep-Convolutional Descriptors. Computer Vision and Pattern Recognition, Boston, 7-12 June 2015, 4305-4314. https://doi.org/10.1109/CVPR.2015.7299059 [8] Murthy, O.V.R. and Goecke, R. (2013) Ordered Trajectories for Large Scale Human Action Recognition. IEEE International Conference on Computer Vision Workshops, Washington DC, 2-8 December 2013, 412-419. https://doi.org/10.1109/ICCVW.2013.61 [9] Shu, T., Xie, D., Rothrock, B., et al. (2015) Joint Inference of groups, Events and Human Roles in Aerial Videos. The IEEE Conference on Computer Vision and Pattern Recognition, Boston, 7-12 June 2015, 4576-4584. [10] Cui, X., Liu, Q., Gao, M., et al. (2011) Abnormal Detection Using Interaction Energy Potentials. The IEEE Conference on Computer Vision and Pattern Recognition, Providence, 20-25 June 2011, 3161-3167. https://doi.org/10.1109/CVPR.2011.5995558 [11] Wu, S., Moore, B.E. and Shah, M. (2010) Chaotic Invariants of La-grangian Particle Trajectories for Anomaly Detection in Crowded Scenes. IEEE Conference on Computer Vision & Pattern Recognition, San Francisco, 13-18 June 2010, 2054-2060. https://doi.org/10.1109/CVPR.2010.5539882 [12] Guo, Z., Li, N., Xu, D., et al. (2013) A Novel Statistical Learning-Based Framework for Automatic Anomaly Detection and Localization in Crowds. IEEE International Conference on Robotics and Biomimetics, Shenzhen, 12-14 December 2013, 1211-1215. https://doi.org/10.1109/ROBIO.2013.6739629 [13] Sabokrou, M., Fathy, M., Hoseini, M., et al. (2015) Real-Time Anomaly Detection and Localization in Crowded Scenes. Computer Vision and Pattern Recognition, Boston, 7-12 June 2015, 56-62. [14] Reddy, V., Sanderson, C. and Lovell, B.C. (2011) Improved Anomaly Detection in Crowded Scenes via Cell-Based Analysis of Foreground Speed, Size and Texture. IEEE Computer Society Conference on Computer Vision and Pattern Recogni-tion Workshops, Colorado Springs, 20-25 June 2011, 55-61. [15] Kratz, L. and Nishino, K. (2009) Anomaly Detection in Ex-tremely Crowded Scenes Using Spatio-Temporal Motion Pattern Models. IEEE Conference on Computer Vision & Pattern Recognition, Miami, 20-25 June 2009, 1446-1453. https://doi.org/10.1109/CVPR.2009.5206771 [16] Feng, J., Zhang, C. and Hao, P. (2010) Online Learning with Self-Organizing Maps for Anomaly Detection in Crowd Scenes. International Conference on Pattern Recognition, Istanbul, 23-26 August 2010, 3599-3602. https://doi.org/10.1109/ICPR.2010.878 [17] Cong, Y., Yuan, J. and Liu, J. (2011) Sparse Reconstruction Cost for Abnormal Event Detection. IEEE Conference on Computer Vision and Pattern Recognition, Providence, 20-25 June 2011, 3449-3456. https://doi.org/10.1109/CVPR.2011.5995434 [18] Ryan, D., Denman, S., Fookes, C., et al. (2011) Textures of Optical Flow for Real-Time Anomaly Detection in Crowds. IEEE International Conference on Advanced Video & Signal-Based Surveillance, Klagenfurt, 30 August-2 September 2011, 1-6. https://doi.org/10.1109/AVSS.2011.6027327 [19] Roshtkhari, M.J. and Levine, M.D. (2013) An On-Line, Real-Time Learning Method for Detecting Anomalies in Videos Using Spatio-Temporal Compositions. Computer Vision & Image Understanding, 117, 1436-1452. https://doi.org/10.1016/j.cviu.2013.06.007 [20] Xu, D., Wu, X., Song, D., et al. (2013) Hierarchical Activity Discovery within Spatio-Temporal Context for Video Anomaly Detection. IEEE International Conference on Image Processing, Melbourne, 15-18 September 2013, 3597-3601. https://doi.org/10.1109/ICIP.2013.6738742 [21] Mehran, R., Oyama, A. and Shah, M. (2009) Abnormal Crowd Behavior Detection Using Social Force Model. Computer Vision and Pattern Recognition, Miami, 20-25 June 2009, 935-942. [22] Zhao, B., Li, F.F. and Xing, E.P. (2011) Online Detection of Unusual Events in Videos via Dynamic Sparse Coding. IEEE Conference on Computer Vision and Pattern Recognition, Providence, 20-25 June 2011, 3313-3320. https://doi.org/10.1109/CVPR.2011.5995524 [23] Lu, C., Shi, J. and Jia, J. (2013) Abnormal Event Detection at 150 FPS in MATLAB. IEEE International Conference on Computer Vision, Sydney, 1-8 December 2013, 2720-2727. https://doi.org/10.1109/ICCV.2013.338 [24] Li, H., Zhang, Y., Yang, M., et al. (2014) A Rapid Abnormal Event Detection Method for Surveillance Video Based on a Novel Feature in Compressed Domain of HEVC. IEEE International Conference on Multimedia and Expo, Chengdu, 14-18 July 2014, 1-6. https://doi.org/10.1109/ICME.2014.6890212 [25] Wang, T. and Snoussi, H. (2015) Detection of Abnormal Events via Optical Flow Feature Analysis. Sensors, 15, 7156-7171. https://doi.org/10.3390/s150407156 [26] Fang, Z., Fei, F., Fang, Y., et al. (2016) Abnormal Event Detection in Crowded Scenes Based on Deep Learning. Multimedia Tools & Applications, 75, 14617-14639. https://doi.org/10.1007/s11042-016-3316-3 [27] Roshtkhari, M.J. and Levine, M.D. (2013) Online Dominant and Anomalous Behavior Detection in Videos. IEEE Conference on Computer Vision & Pattern Recognition, Portland, 23-28 June 2013, 2611-2618. https://doi.org/10.1109/CVPR.2013.337 [28] Mahadevan, V., Li, W., Bhalodia, V., et al. (2010) Anomaly Detection in Crowded Scenes. IEEE Conference on Computer Vision and Pattern Recognition, San Francisco, 13-18 June 2010, 1975-1981. https://doi.org/10.1109/CVPR.2010.5539872 [29] Li, W., Mahadevan, V. and Vasconcelos, N. (2014) Anomaly Detection and Localization in Crowded Scenes. IEEE Transactions on Pattern Analysis & Machine Intelligence, 36, 18-32. https://doi.org/10.1109/TPAMI.2013.111 [30] Adam, A., Rivlin, E., Shimshoni, I., et al. (2007) Robust Real-Time Unusual Event Detection Using Multiple Fixed-Location Monitors. IEEE Transactions on Pattern Analysis & Machine Intelligence, 30, 555-560. https://doi.org/10.1109/TPAMI.2007.70825 [31] Cong, Y., Yuan, J. and Liu, J. (2013) Abnormal Event Detection in Crowded Scenes Using Sparse Representation. Pattern Recognition, 46, 1851-1864. https://doi.org/10.1016/j.patcog.2012.11.021