Objects matching algorithm of multi-camera across non-overlapping views
Chen Wei · Journal of Computer Applications · 2010
Multiple non-overlapping cameras are required in visual surveillance of wide areas,and the observations of objects are often widely separated in time and space.The key problem of multiple-camera objects tracking is how to match the same object in different cameras.A objects matching algorithm based on space-time and appearance model relationship was proposed in this paper.The algorithm learned the inter-camera relationship by kernel density estimation to count multivariate probability density of space-time variables,and used Brightness Transfer Function(BTF) from one camera to another to handle the appearance model change of an object as it moves from one camera to another.Experiments with real videos and simulation validate the proposed approach.