A track-before-detect algorithm with successive track cancellation

In this work we introduce a novel track formation procedure to overcome the performance losses of the two-stage track-before-detect architecture in [1] when targets are closely-spaced. The proposed strategy adopts a dynamic programming algorithm which successively estimates the trajectories of prospective targets and removes the corresponding measurements from the input data, and which does not need a discretization of the state space. Experimental results demonstrate that its detection capabilities are almost independent of the mutual distance of the targets and equal to those of the system in [1] when targets are widely-spaced.