A filter for distinguishable and independent populations
Emmanuel Delande, Jérémie Houssineau, Daniel E. Clark · arXiv (Cornell University) · 2015
This article introduces a multi-object filter for the resolution of joint detection/tracking problems involving multiple targets, derived from the novel Bayesian estimation framework for stochastic populations. Fully probabilistic in nature, the filter for Distinguishable and Independent Stochastic Populations (DISP) exploits two exclusive probabilistic representations for the potential targets. The distinguishable targets are those for which individual information is available through past detections; they are represented by individual tracks. The indistinguishable targets are those for which no individual information is available yet; they are represented collectively by a single stochastic population. Assuming that targets are independent, and adopting the most one measurement per scan per target rule, the DISP filter propagates the set of all possible tracks, with associated credibility, based on the sequence of measurement sets collected by the sensor so far. A few filtering approximations, aiming at curtailing the computational cost of a practical implementation, are also discussed.