Online Zoom Selection Approaches for Coverage Redundancy in Visual Sensor Networks

When a network of cameras with adjustable zoom lenses is tasked with object coverage, an important question is how to determine the optimal zoom level for each camera. While covering a smaller area allows for higher detection likelihood, overlapping fields of view introduce a redundancy which is vital to fault tolerance and acquisition of multiple perspectives of targets. In this paper, we study the coverage redundancy problem in visual sensor networks formalised as the k-coverage. We propose a resolution-based detection model that enables the exploration of the zoom level impact on the extension of coverage redundancy. Given mobile targets, we investigate the network-wide best k-coverage using global knowledge of the network. We explore the advantages and disadvantages of this approach and propose a realistic zoom adaptation model for cameras under a new zoom level proximity constraint. Further to the global approach, which aim to provide the highest possible k-coverage across the network, we study the performance of several online heuristics and learning approaches which use only local knowledge to approximate the highest k-coverage under our zoom level proximity constraint. Furthermore, we show that in many scenarios the dynamic behaviour resulting from the online learning approaches is not only computationally less expensive but also leads to a significant improvement on the level of coverage redundancy compared with heuristic approaches. This becomes even more apparent when objects follow scripted movement patterns.