A socio-economic approach to online vision graph generation and handover in distributed smart camera networks

Abstract

Abstract-In this paper we propose an approach based on selfinterested autonomous cameras, which exchange responsibility for tracking objects in a market mechanism, in order to maximise their own utility. A novel ant-colony inspired mechanism is used to grow the vision graph during runtime, which may then be used to optimise communication between cameras. The key benefits of our completely decentralised approach are on the one hand generating the vision graph online which permits the addition and removal cameras to the network during runtime and on the other hand relying only on local information, increasing the robustness of the system. Since our market-based approach does not rely on a priori topology information, the need for any multicamera calibration can be avoided.

The Future of Camera Networks

“…They use auctions to form coalitions of cameras to achieve specific tasks. Similarly, Esterle et al [10,13] use auctions to find neighbouring cameras for continuous tracking. They introduce artificial pheromones to adapt to changes in the network.…”

“…This ensures efficiency of network-wide resources, but requires either perfect re-identification or knowledge of the environment to ensure sufficient re-identification of targets among different cameras [10,13,31,34,48].…”

Online Multi-object k-coverage with Mobile Smart Cameras

“…Our work is based on the ideas proposed by Esterle et al [1,3], who introduced a market-based approach in combination with arti cial pheromones to e ciently coordinate tracking tasks in smart camera networks with limited resources. We transfer these ideas to networks of mobile smart cameras and the problem of distributed k-coverage.…”

“…We model this information as a graph, that can be interpreted as a neighbourhood relation between the two cameras. Inspired by the pheromones used in the foraging process of ants and the work by Esterle et al [1,3] on static smart camera networks, an edge is reinforced when two cameras observe the same object. At each discrete time step when a camera c i observes a common object o a with another camera c j , we increase the value on a local link λ(i, j).…”

“…In networks of static smart cameras, Esterle et al [1,3] showed how the e ciency of decentralised coordination (in terms of task performance and resource overhead) could be improved by cameras learning the neighbourhood relations between their elds of view (FoVs) at runtime. Here, we ask if this might aid the coordination of a network of mobile cameras.…”