基于时空特征点的群体异常行为检测算法

被引:6
作者
王传旭
董晨晨
机构
[1] 青岛科技大学信息学院
关键词
群体异常行为; 时空特征点; 关键词词袋; 高斯混合模型;
D O I
10.16337/j.1004-9037.2012.04.015
中图分类号
TP391.41 [];
学科分类号
080203 ;
摘要
提出了用时空特征点描述群体行为的新方法。首先对比分析时空Harris角点、Gabor小波、Hessian矩阵三种特征点提取方法,选择了基于Hessian矩阵的尺度不变方法提取特征点;分别采用梯度直方图、光流直方图以及时空Haar特征三种方法对特征点构建描述符。采用Bag-of-words策略对正常行为建模,使用基于EM估计的高斯混合模型建模产生关键词,根据关键词为每一视频片段建立一个带有概率分布的编码向量,形成编码表。异常行为的检测是将测试样本的编码向量与训练样本编码表进行比较,计算相似度距离,当最小距离大于阈值时,判该群体行为异常。在UCF和UMN两种群体行为数据集下的实验结果表明,该方法能够对群体异常行为进行有效识别,对尺度变化以及背景光照变化等具有较好的适应性。
引用
收藏
页码:422 / 428
页数:7
相关论文
共 12 条
[1]  
A survey on vision-based human action recognition. Ronald Poppe. Image and Vision Computing . 2010
[2]  
Abnormal Crowd Behavior Detection using Social Force Model. Ramin Mehran,Alexis Oyama,Mubarak Shah. IEEE Conference on Computer Vision and Pattern Recognition . 2009
[3]  
Anomaly Detectionin CrowdedScenes. Vijay Mahadevan,Weixin Li,Viral Bhalodia,Nuno Vasconcelos. EEE Conf. on Computer Vision and Pattern Recognition . 2010
[4]  
Modeling, Clustering, and Segmenting Video with Mixtures of Dynamic Textures. A.B.Chan,N.Vasconcelos. IEEE Transactions on Pattern Analysis and Machine Intelligence . 2008
[5]  
Behavior recognition via sparse spatio-temporal features. P. Dollar,V. Rabaud,G. Cottrell,S. Belongie. Proceedings of IEEE International Workshop on Visual Surveillance and Performance Evaluation of Tracking and Surveillance . 2005
[6]  
3DHaar-like features for pedestrian detection. Cui Xinyi,Liu Yazhou,Shan Shiguang,et al. IEEE International Conference on Multimedia andExpo . 2007
[7]  
Chaoticinvariants of Lagrangian particle trajectories foranomaly detection in crowded scenes. Wu Shandong,Moore B E,Shah Mubarak. IEEEComputer Society Conference on Computer Visionand Pattern Recognition (CVPR) . 2010
[8]  
Space-time Interest Points. I. Laptev,T. Lindeberg. Proceedings of IEEE International Conference on Computer Vision . 2003
[9]  
An Efficient Dense and Scale-Invariant Spatio-Temporal Interest Point Detector. G. Willems,T. Tuytelaars,L. V. Gool. ECCV . 2008
[10]  
A Streakline Representation of Flow in Crowded Scenes. Ramin Mehran,Brian E. Moore,Mubarak Shah. ECCV 2010, Part III .