Due to the broad coverage of an environment and the possibility of coordination among different cameras, video sensor networks have attracted much interest in recent years. Although the field-of-view (FOV) of a single camera is limited and cameras may have overlapping/non-overlapping FOVs, seamless tracking of moving objects is desired. As the increasing of the video network complexity, there are more and more camera nodes in a network. This makes it hard for human observers to take care of the entire system and brings the emergence of the camera selection, handoff and control technologies.
In this study, we introduce a series of economics frameworks into the camera selection, handoff and control problem. This starts with two game theoretical approaches - the potential game approach and the weakly acyclic game approach. With these two methods, we can model the camera selection and handoff problem as a multiplayer game. Existing learning algorithms in the game theory literature make it efficient to find an optimal as well as stable solution to this problem.
As camera selection and handoff largely depend on the accuracy of the applied trackers, we develop a technique to jointly consider the tracking problem and the camera selection problem. In this work, fusion of multiple trackers is integrated with the camera selection process in a closed-loop manner.
Finally, active camera controls are considered by using the auction protocol. Unlike previous work, the bid price is formulated to have a vector representation, such that when a camera is available to follow multiple objects, we consider the "willingness" of this camera to track a particular object. Meanwhile, the potentially available cameras can also be considered to follow an object after some panning or tilting operations. Most of the computation is decentralized by computing the bid price locally while the final assignment is made by a virtual auctioneer based on all the available bids, which is analogous to a real auction in economics. Thus, we can take the advantage of distributed/centralized computation and avoid their pitfalls.
All these approaches are evaluated with real-world data under the VideoWeb  camera network environment. These proposed approaches are also compared with each other and other approaches. The experimental results show the robustness and efficacy of this study.