Finite Point Process Approach to Multi-Target Localization.

摘要

A finite point process approach to multi-target localization from a transient signal is presented. After modeling the measurements as a Poisson point process, we propose a twofold scheme that includes an expectation maximization algorithm to estimate the target locations for a given number of targets and an information theoretic algorithm to select the target model, i.e., number of targets. Similar to the finite point process solution for the multi- target tracking, i.e., the probability hypothesis density filter, the proposed localization scheme does not require solving the data association problem and can account for clutter noise as well as missed detection. The optimal subpattern assignment metric is used to assess the performance and accuracy of the proposed localization algorithm. Implementation of the proposed algorithm on synthetic data yields desirable results. The proposed algorithm is then applied to multi-shooter localization problem using acoustic gunfire detection systems.