Consensus in visual sensor networks consisting of calibrated and uncalibrated cameras

Fusing multiple views of different cameras in a visual sensor network (VSN) require the calculation of a consensus. As a result we are able to overcome the uncertainties from the tracking algorithm and physical constraints in the observed environment. This paper addresses the topic of reaching a consensus within VSN of calibrated or uncalibrated cameras. In contrast to several proposed techniques, where the consensus is calculated on all cameras simultaneously, our approach calculates the consensus on a single camera. Furthermore, the responsibility for this calculation is migrated to the camera actually detecting the object with a better subjective view than its neighbors. This saves resources and reduces the communication overhead within the network.