A Framework for Tracking Moving Target in a Heterogeneous Camera Suite

This paper presents a novel approach for tracking a moving target of interest across a large-scale distributed camera system where some of the cameras are mobile. The growing trend of fitting public transport vehicles like buses, trains and taxis with some form of surveillance system, mainly to counter vandalism has created the opportunity for these systems to be integrated to a larger network-based surveillance system that has the potential to span the entire city. Specifically, we address the problem of using mobile surveillance cameras to help increase the visual footprint of existing fixed surveillance infrastructure. We introduce a grid-based Bayesian filtering approach together with a map to systematically combine the multiple observations coming from the heterogeneous surveillance system. Experimental evidences have indicated that our approach has the potential to deal with large-scale observations that are prevalent in city-wide implementation