Apprentissage de dictionnaire pour un modèle d'apparence parcimonieux en suivi visuel

Abstract : This paper presents a novel approach to visual object tracking based on particle filtering. The appearance of the target object is described by a sparse representation provided by dictionary learning, which leads to a model of reduced dimension. The likelihood of a candidate region is built on a similarity measure which can be interpreted as the result of a matched filter in the new representation space formed by the dictionary. Thus it can optimally detect a set of reference patches extracted from the target at known positions in the candidate region. Experimental validation shows the efficiency and the robustness of the proposed approach.
Document type :
Conference papers
Liste complète des métadonnées

Cited literature [8 references]  Display  Hide  Download

https://hal.archives-ouvertes.fr/hal-01211262
Contributor : Sylvain Rousseau <>
Submitted on : Sunday, October 4, 2015 - 2:40:11 PM
Last modification on : Saturday, March 23, 2019 - 1:25:08 AM
Document(s) archivé(s) le : Tuesday, January 5, 2016 - 10:13:37 AM

File

master.pdf
Files produced by the author(s)

Identifiers

  • HAL Id : hal-01211262, version 1

Citation

Sylvain Rousseau, Christelle Garnier, Pierre Chainais. Apprentissage de dictionnaire pour un modèle d'apparence parcimonieux en suivi visuel. GRETSI, Sep 2015, Lyon, France. ⟨hal-01211262⟩

Share

Metrics

Record views

194

Files downloads

205