Abstract
Data-driven methods of simulating a crowd of virtual humans that exhibit behaviors imitating real human crowds play an important role in crowd simulation. In this paper, we propose a Bayesian framework for the extraction of real human's behaviors which exhibit interactions in their daily life using multiple fixed cameras. The described Markov chain Monte Carlo particle filter can effectively deals with interacting targets which are influenced by the proximity and behaviors of other targets. In this paper, we use a Markov random field motion prior combing with a federate filter algorithm which treats the observations discriminatorily to substantially improve the tracking of a fixed number of interacting targets. Simultaneously, we replace the traditional importance sampling step with MCMC sampling step to get over the vast computational requirements for large numbers of targets. i.e., we focus on the data fusion and the behavior recognition process. Finally, experimental results demonstrate that the proposed Bayesian framework deals efficiently and effectively with extractions of interacting behavior.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have