Relations as Context to Improve Multi-Target Tracking and Activity Recognition

The explicit recognition of the relationships between interacting objects can improve the understanding of their dynamic model. In this work, we investigate the use of Relational Dynamic Bayesian Networks to represent the dependencies between objects behavior in the context of multitarget tracking. We propose a new formulation of the transition model that accommodates for First-Order Logic relations and we extend the Particle Filter algorithm in order to directly track relations between targets. Many applications can benefit from this work, as activities recognition, traffic monitoring, strategic analysis, sports coaching and others. We present some results about activity recognition in monitoring Canadian costal borders.