Spatial-temporal activity interactions detection in video survalliance

In this paper, a novel framework to explore the activity spatial-temporal interactions in complex video surveillance scenes is proposed. Firstly, low-level motion features are detected and quantized into words. The Hierarchical Dirichlet Processes model is then applied to automatically cluster low-level features into atomic activities. Afterwards, the dynamic behaviors of the activities are represented as a multivariate point-process. The pair-wise causal scores and periods between activities are explicitly captured by the non-parametric Granger causality analysis, from which the activity spatial-temporal interactions are discovered. The results of the real world traffic datasets demonstrate the effectiveness of the proposed method.