Multitarget-multisensor ML and PHD: Some asymptotics

Multi-object estimation transforms unlabeled (and possible false and missing) observations to estimates that are also unlabeled. Two estimation strategies are here studied. The first one is a two-step procedure: detection of the number of objects followed by estimation of their locations. The second one appeals to Random Finite Set (RFS) theory, and is based on the Probability Hypothesis Density (PHD). This paper proves the notion that both are asymptotically efficient, thus achieving the same performance of a clairvoyant (in number of objects) scheme.