基于时空信息约束的密集人群分割方法

doi:10.12066/j.issn.1007-2861.1760

上海大学学报(自然科学版) ›› 2017, Vol. 23 ›› Issue (5): 742-751.doi: 10.12066/j.issn.1007-2861.1760

基于时空信息约束的密集人群分割方法

易娴^1,2, 董楠², 魏建明², 朱文浩¹

1. 上海大学计算机工程与科学学院, 上海 200444;
2. 中国科学院上海高等研究院安全与应急实验室, 上海 201210

收稿日期:2015-11-02 出版日期:2017-10-30 发布日期:2017-10-30
通讯作者: 董楠(1982—), 男, 副研究员, 博士, 研究方向为视频图像理解、机器学习等. E-mail: dongn@sari.ac.cn
作者简介:董楠(1982—), 男, 副研究员, 博士, 研究方向为视频图像理解、机器学习等. E-mail: dongn@sari.ac.cn
基金资助:
国家自然科学基金资助项目(61302151); 上海市自然科学基金资助项目(13ZR1455100)

Segmenting groups in crowd based on spatiotemporal restraint

YI Xian^1,2, DONG Nan², WEI Jianming², ZHU Wenhao¹

1. School of Computer Engineering and Science, Shanghai University, Shanghai 200444, China;
2. Safety and Emergency Laboratory, Shanghai Advanced Research Institute, Chinese Academy of Sciences, Shanghai 201210, China

Received:2015-11-02 Online:2017-10-30 Published:2017-10-30

摘要/Abstract

摘要：

在密集场景中, 人群可以看作是由三五成群的、具有运动一致性的子群体构成的集合. 为了能划分这些子群体, 以便对人群的交互行为进行分析, 提出了一种基于时空信息约束的分割方法. 该方法首先利用背景建模和特征点跟踪, 获取视频图像帧中运动目标的时空信息; 然后, 利用前景中群体分布的空间区域信息, 将空间上相邻近的人群划分为若干个子群体; 最后, 通过一段时间内的运动相关性将子群体进行进一步的分割. 这两种约束信息相互作用, 得到具有运动一致性的子群体. 通过大量实验验证测试, 表明该方法可以应用于不同人群密度、视角范围等多种场景视频.

关键词: 时空信息约束, 运动一致性, 子群体, 群体分割

Abstract:

Pedestrians in crowded scenes can be seen as a collection of groups moving coherently. To segment the crowd into groups and analyze interactions between them, a method using group spatiotemporal relationships is proposed. Spatiotemporal information of crowds by modeling the background and tracking the feature points is acquired. Then the individuals are grouped according to aspatial neighborhood restraint. The groups are further divided by motion correlation over time. These two restraints work with each other effectively, and generate groups with coherent motion. Tests on many videos of real-world pedestrian scenes show that the method can be applied to a variety of scenes with different crowd densities and perspective of videos.

Key words: coherent motion, spatiotemporal restraint, sub-group, crowd segmentation

易娴1,2, 董楠2, 魏建明2, 朱文浩1. 基于时空信息约束的密集人群分割方法[J]. 上海大学学报(自然科学版), 2017, 23(5): 742-751.

YI Xian1,2, DONG Nan2, WEI Jianming2, ZHU Wenhao1. Segmenting groups in crowd based on spatiotemporal restraint[J]. Journal of Shanghai University（Natural Science Edition）, 2017, 23(5): 742-751.

参考文献

[1] Ben X Y, Huang X F, Zhuang Z Y, et al. Agent-based approach for crowded pedestrian evacuation simulation [J]. Intelligent Transport Systems, 2013, 7(1): 55-67.
[2] Li J Y, Lu X F, Ding L Y, et al. Moving target tracking via particle filter based on color and contour features [C]// Proceedings of International Conference on Information Engineering and
Computer Science. 2010: 1-4.
[3] Zhang Q, Rui T, Fang H S, et al. Particle filter object tracking based on Harris-SIFT feature matching [J]. Procedia Engineering, 2012, 29: 924-929.
[4] Andrade E L, Blunsden S, Fisher R B. Modeling crowd scenes for event detection [C]//Proceedings of 18th International Conference on Pattern Recognition. 2006: 175-178.
[5] Andrade E L, Fisher R B. Simulation of crowd problems for computer vision [C]// First International Workshop on Crowd Simulation. 2005: 71-80.
[6] Chan A, Vasconcelos N. Modeling, clustering, and segmenting video with mixtures of dynamic textures [J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2008, 30(5): 909-926.
[7] Aveni A. The not-so-lonely crowd: friendship groups in collective behavior [J]. Sociometry, 1977, 40: 96-99.
[8] Karamouzas I, Overmars M. Simulating and evaluating the local behavior of small pedestrian groups [J]. IEEE Transactions on Visualization and Computer Graphics, 2012, 18(3): 394-406.
[9] Lan T, Wang Y, Yang W L, et al. Discriminative latent models for recognizing contextual group activities [J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2012,
34(8): 1549-1562.
[10] Moussaid M, Perozo N, Garnier S, et al. The walking behavior of pedestrian social groups and its impact on crowd dynamics [J]. PloS One, 2010, 5(4): e10047.
[11] Chang M C, Krahnstoever N, Lim S, et al. Group level activity recognition in crowded environments across multiple cameras advanced video and signal based surveillance [C]// Proceedings of International Conference on Advanced Video and Signal Based Surveillance. 2010: 56-63.
[12] Ge W N, Collins R T, Ruback B. Automatically detecting the small group structure of a crowd [C]// Proceedings of Workshop on Applications on Computer Vision. 2009: 1-8.
[13] Kaewtrakulpong P, Bowden R. An improved adaptive background mixture model for realtime tracking with shadow detection [C]// Proceedings of 2nd European Workshop on Advanced Video Based Surveillance Systems. 2001: 149-158.
[14] Tomasi C, Kanade T. Detection and tracking of point features [J]. International Journal of Computer Vision, 1991, 9: 137-154.
[15] Shao J. CUHK crowd dataset [EB/OL]. [2015-04-30]. http://www.ee.cuhk.hk/jshao/cuHKcrowdfiles/cuhk-dataset-htm.
[16] Shao J, Chen C L, Wang X G. Scene-independent group profiling in crowd [C]// Proceedings of Conference on Computer Vision and Pattern Recognition. 2014: 2227-2234.
[17] Zhou B, Tang X, Wang X. Coherent filtering: detecting coherent motions from crowd clutters [C]// Proceedings of Conference on Computer Vision. 2012: 857-871.
[18] Barnich O, Van Droogenbroeck M. Vibe: a universal background subtraction algorithm for video sequences [J]. IEEE Transactions on Image Processing, 2011, 20(6): 1709-1724.

基于时空信息约束的密集人群分割方法

Segmenting groups in crowd based on spatiotemporal restraint

PDF

可视化

摘要/Abstract

引用本文

使用本文

参考文献

相关文章 0

编辑推荐

Metrics

本文评价