On data association with possibly unresolved measurements - Archive ouverte HAL Accéder directement au contenu
Communication Dans Un Congrès Année : 2023

On data association with possibly unresolved measurements

Résumé

Tracking targets based on measurements provided by radar, sonar, or lidar sensors is essential to obtain situational awareness in important applications, including autonomous navigation and applied ocean sciences. A key challenge in multitarget tracking is the unknown association between the available measurements and the targets to be tracked. In particular, robust data association for closely spaced targets requires advanced methods that explicitly model unresolved measurements. Due to limited sensor resolution, the sensor produces a single measurement for two or more actual targets. If not explicitly modeled in the multitarget tracking method, unresolved measurements lead to track losses and thus, to significant tracking errors. In this paper, we propose a scalable data association method for the tracking of multiple potentially unresolved targets. A loopy belief propagation method is presented that efficiently approximates the marginal association probabilities given a set of potentially unresolved measurements. This method scales quadratically in the number of targets and linearly in the number of measurements. Our numerical results demonstrate that the computed approximate marginal association probabilities are close in L1 distance to the true marginal association probabilities, which can only be calculated for very small tracking scenarios.
Fichier sous embargo
Fichier sous embargo
0 1 9
Année Mois Jours
Avant la publication
mercredi 5 juin 2024
Fichier sous embargo
mercredi 5 juin 2024
Connectez-vous pour demander l'accès au fichier

Dates et versions

hal-04325357 , version 1 (05-12-2023)

Licence

Copyright (Tous droits réservés)

Identifiants

Citer

Augustin A Saucan, Florian Meyer. On data association with possibly unresolved measurements. 2023 26th International Conference on Information Fusion (FUSION), Jun 2023, Charleston, United States. pp.1-8, ⟨10.23919/FUSION52260.2023.10224130⟩. ⟨hal-04325357⟩
22 Consultations
1 Téléchargements

Altmetric

Partager

Gmail Facebook X LinkedIn More