Abstract
This paper presents a scalable solution to the problem of tracking objects across spatially separated, uncalibrated, non-overlapping cameras. Unlike other approaches this technique uses an incremental learning method, to model both the colour variations and posterior probability distributions of spatio-temporal links between cameras. These operate in parallel and are then used with an appearance model of the object to track across spatially separated cameras. The approach requires no pre-calibration or batch preprocessing, is completely unsupervised, and becomes more accurate over time as evidence is accumulated.