Decoupled Multicamera Sensing for Flexible View Generation

Abstract

Any sensing paradigm has three important components, namely, theactor, thesensor, and theenvironment. Traditionally, the sensors have been attached to either the actor or the environment. This restricts the kind of sensing that can be undertaken. We study a newer decoupled sensing paradigm, which separates the sensors from both the actor and the environment and tremendously increases the flexibility with which the scenes can be viewed. For example, instead of showing just one view, “how the environment sees the actor” or “how the actor sees the environment,” a viewer can choose to see either one or both of these views and even choose to see the scene from any desired position in any desired direction. We describe a methodology using mobile autonomous sensors to undertake such decoupled sensing and study the feasible number as well as the placement of such sensors. Also, we describe how the sensors can coordinate their movements around a moving actor so as to continue capturing the required views with minimum overall cost. The practical results obtained demonstrate the viability of the proposed approach.