Multiple Target Tracking via Multimarginal Optimal Transport

Abstract: The field of Multiple Target Tracking (MTT) deals with the task of estimating targets that appear, disappear, and move within a scene, given data from noisy measurements. An important step in an MTT algorithm is data association: determining if a detection is a false positive, and if not, which target it should correspond to. We find that this task is closely related to the field of multimarginal optimal transport (MMOT), which originally stems from the problem of optimally moving dirt for the construction of roads. In this work, we connect MTT with MMOT. Specifically, we show how a popular evaluation metric for MTT (the TGOSPA metric) can be cast as an MMOT problem, allowing it to be computed efficiently using techniques from the MMOT literature.

numerical analysisoptimization and control

Audience: researchers in the topic

( paper )

CAM seminar

Series comments: Online streaming via zoom on exceptional cases if requested. Please contact the organizers at the latest Monday 11:45.