Compressive measurement for target tracking in persistent, pervasive surveillance applications

摘要

Motion tracking in persistent surveillance applications enters an interesting regime when the movers are of a size on the order of the image resolution elements or smaller. In this case, for reasonable scenes, information about the movers is a natively sparse signal--in an observation of a scene at two closely separated time-steps, only a small number of locations (those associated with the movers) will have changed dramatically. Thus, this particular application is well-suited for compressive sensing techniques that attempt to efficiently measure sparse signals. Recently, we have been investigating two different approaches to compressive measurement for this application. The first, differential Combinatorial Group Testing (dCGT), is a natural extension of group testing ideas to situations where signal differences are sparse. The second methodology is an l-1-minimization based recovery approach centered on recent work in random (and designed) multiplex sensing. In this manuscript we will discuss these methods as they apply to the motion tracking problem, discuss various performance limits, present early simulation results, and discuss notional optical architectures for implementing a compressive measurement scheme.