ABSTRACT
This chapter proposes a multisensor multitarget Classification architecture. Each sensor is supposed to track locally a set of randomly appearing and disappearing targets using interacting multiple models (IMM) algorithms. The chapter provides some basics on belief function theory. It also provides some information on the adopted tracking solution. The chapter describes the Bayesian and the credal classifiers that are locally processed at each sensor level. It highlights the two main contributions: the parameterless track-to-track solution and the motivation of using the disjunctive credal rule of combination, which are processed at the fusion center. The chapter presents a nested class simulation example that describes piracy surveillance where the proposed fusion solution is shown to be efficient, especially in the case of high sensor uncertainties. It explains the idea behind multitarget tracking problem and provides some information about the adopted solution to estimate trajectories of targets.
