Track-before-detect procedures in multi-targets environments

In this paper, track-before-detect (TBD) procedures for radar systems are presented and discussed. After introducing the signal model we rely on, the non-Bayesian approach developed in [2,3] with reference to a single-target scenario is extended to multiple targets and an optimum (for discretized statistics) TBD algorithm is derived. Interestingly, this algorithm admits a Viterbi-like implementation with a complexity linear in the number of integrated frames, as in the single-target case; however, the number of states to be considered in the trellis-diagram now grows exponentially with the number of targets. Successively, two sub-optimum TBD procedures are investigated, which allow to trade better estimation and tracking accuracy for lower implementation complexity. Finally, numerical examples are provided to assess and compare the performance of the proposed TBD algorithms.