Sub-sample swapping for sequential Monte Carlo approximation of high-dimensional densities in the context of complex object tracking

Severine Dubuisson 1 Christophe Gonzales 2 Xuan Son Nguyen 2
1 PEQUAN - Performance et Qualité des Algorithmes Numériques
LIP6 - Laboratoire d'Informatique de Paris 6
2 DECISION
LIP6 - Laboratoire d'Informatique de Paris 6
Abstract : In this paper, we address the problem of complex object tracking using the par- ticle lter framework, which essentially amounts to estimate high-dimensional distributions by a sequential Monte Carlo algorithm. For this purpose, we rst exploit Dynamic Bayesian Networks to determine conditionally independent subspaces of the object's state space, which allows us to independently perform the particle lter's propagations and corrections over small spaces. Second, we propose a swapping process to transform the weighted particle set provided by the update step of the particle lter into a \new particle set" better focus- ing on high peaks of the posterior distribution. This new methodology, called Swapping-Based Partitioned Sampling, is proved to be mathematically sound and is successfully tested and validated on synthetic video sequences for single or multiple articulated object tracking.
Document type :
Journal articles
Complete list of metadatas

https://hal.sorbonne-universite.fr/hal-00820473
Contributor : Xuan Son Nguyen <>
Submitted on : Monday, May 13, 2013 - 9:36:39 AM
Last modification on : Thursday, March 21, 2019 - 1:10:41 PM
Long-term archiving on : Monday, August 19, 2013 - 3:10:08 PM

File

ijar.pdf
Files produced by the author(s)

Identifiers

Citation

Severine Dubuisson, Christophe Gonzales, Xuan Son Nguyen. Sub-sample swapping for sequential Monte Carlo approximation of high-dimensional densities in the context of complex object tracking. International Journal of Approximate Reasoning, Elsevier, 2013, Special issue: Uncertainty in Artificial Intelligence and Databases, 54 (7), pp.934-953. ⟨10.1016/j.ijar.2013.03.002⟩. ⟨hal-00820473⟩

Share

Metrics

Record views

217

Files downloads

462