Advances in Data Fusion Architectures
Stefano P. Coraluppi, Craig A. Carthel · 2014
In this chapter, the author discusses a number of settings where multistage processing is beneficial. These are the following: single-sensor dense-target scenarios; multiscale sensor scenarios; large networks of low-performance sensors; and multiscale object scenarios. The chapter focuses on the interest to tracking and fusion practitioners whether or not an multiple-hypothesis tracking (MHT) paradigm is adopted. Interestingly, the authors find that the usual assumption is estimation and detection theory does not hold in target tracking. The reason is that optimal tracking is intractable; thus the authors are faced with either suboptimal centralized single-stage processing or with judiciously chosen alternatives. The unifying philosophy in the examples is that tracking at reduced complexity often can be achieved by appropriate concatenation of relatively simple processing stages.