Discovering Spatio-Temporal Relationships Among Activities in Videos Using a Relational Topic-Transition Model

Discovering motion activities in videos is a key problem in computer vision, with applications in scene analysis, video categorization, and video indexing. In this paper, we propose a method that uses probabilistic topic modeling for discovering patterns of motion that occur in a given activity. Our method also identifies how the discovered patterns of motion relate to one another in space and time. The topic-modeling approach used by our method is the relational topic model. Our experiments show that our method is able to discover relevant spatio-temporal motion patterns in videos.

[1]  Mark Steyvers,et al.  Finding scientific topics , 2004, Proceedings of the National Academy of Sciences of the United States of America.

[2]  David M. Blei,et al.  Relational Topic Models for Document Networks , 2009, AISTATS.

[3]  Shaogang Gong,et al.  A Markov Clustering Topic Model for mining behaviour in video , 2009, 2009 IEEE 12th International Conference on Computer Vision.

[4]  Michael I. Jordan,et al.  Latent Dirichlet Allocation , 2001, J. Mach. Learn. Res..

[5]  Chris Stauffer,et al.  Estimating Tracking Sources and Sinks , 2003, 2003 Conference on Computer Vision and Pattern Recognition Workshop.

[6]  Shaogang Gong,et al.  Video Behaviour Mining Using a Dynamic Topic Model , 2011, International Journal of Computer Vision.

[7]  Luc Van Gool,et al.  What's going on? Discovering spatio-temporal dependencies in dynamic scenes , 2010, 2010 IEEE Computer Society Conference on Computer Vision and Pattern Recognition.