A Belief Based Correlated Topic Model for Semantic Region Analysis in Far-Field Video Surveillance Systems

摘要

In this paper, a belief based correlated topic model (BCTM) is proposed for the semantic region analysis of pedestrian motion patterns in the crowded scenes. The inputs of the BCTM can be holistic trajectories or fragments of trajectories. By integrating the sources, sinks, and a forest of randomly spanning trees of trajectories as priors, the proposed BCTM improves the learning of semantic regions, significantly. In addition, the model can also cluster topics through modeling relations among topics. Experiments on a large scale data set, which are collected from the crowded New York Grand Central Station, show that the BCTM outperforms the state-of-the-art methods on qualitative results of learning semantic regions.