A track-before-detect procedure for sparse data

In this paper, we consider the problem of detecting the presence of a prospective, moving target from a set of noisy measurements. We propose a two-steps approach: The first step discards unreliable measurements (i.e., those whose likelihood ratio falls below a preassigned threshold); The second step, instead, exploits the correlation among observations taken at different time instants and makes the final decision. A novel, computationally efficient, track-before-detect algorithm which exploits the sparse nature of the measurements is proposed, and experimental results to asses the algorithm performance are provided.