Multi-target tracking with sparse group features and position using discrete-continuous optimization

Billy Peralta, Alvaro Soto

Resultado de la investigación: Conference contribution

Resumen

Multi-target tracking of pedestrians is a challenging task due to uncertainty about targets, caused mainly by similarity between pedestrians, occlusion over a relatively long time and a cluttered background. A usual scheme for tackling multi-target tracking is to divide it into two sub-problems: data association and trajectory estimation. A reasonable approach is based on joint optimization of a discrete model for data association and a continuous model for trajectory estimation in a Markov Random Field framework. Nonetheless, usual solutions of the data association problem are based only on location information, while the visual information in the images is ignored. Visual features can be useful for associating detections with true targets more reliably, because the targets usually have discriminative features. In this work, we propose a combination of position and visual feature information in a discrete data association model. Moreover, we propose the use of group Lasso regularization in order to improve the identification of particular pedestrians, given that the discriminative regions are associated with particular visual blocks in the image. We find promising results for our approach in terms of precision and robustness when compared with a state-of-the-art method in standard datasets for multi-target pedestrian tracking.

Idioma originalEnglish
Título de la publicación alojadaComputer Vision - 12th Asian Conference on Computer Vision, ACCV 2014, Revised Selected Papers
EditoresC.V. Jawahar, Shiguang Shan
EditorialSpringer Verlag
Páginas680-694
Número de páginas15
ISBN (versión impresa)9783319166339
DOI
EstadoPublished - 1 ene 2015
Evento12th Asian Conference on Computer Vision, ACCV 2014 - Singapore, Singapore
Duración: 1 nov 20142 nov 2014

Serie de la publicación

NombreLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volumen9010
ISSN (versión impresa)0302-9743
ISSN (versión digital)1611-3349

Conference

Conference12th Asian Conference on Computer Vision, ACCV 2014
PaísSingapore
CiudadSingapore
Período1/11/142/11/14

Huella dactilar

Multi-target Tracking
Data Association
Continuous Optimization
Discrete Optimization
Target tracking
Target
Trajectories
Trajectory
Association Model
Lasso
Discrete Data
Discrete Model
Occlusion
Data Model
Random Field
Divides
Regularization
Robustness
Uncertainty
Vision

ASJC Scopus subject areas

  • Theoretical Computer Science
  • Computer Science(all)

Citar esto

Peralta, B., & Soto, A. (2015). Multi-target tracking with sparse group features and position using discrete-continuous optimization. En C. V. Jawahar, & S. Shan (Eds.), Computer Vision - 12th Asian Conference on Computer Vision, ACCV 2014, Revised Selected Papers (pp. 680-694). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 9010). Springer Verlag. https://doi.org/10.1007/978-3-319-16634-6_49
Peralta, Billy ; Soto, Alvaro. / Multi-target tracking with sparse group features and position using discrete-continuous optimization. Computer Vision - 12th Asian Conference on Computer Vision, ACCV 2014, Revised Selected Papers. editor / C.V. Jawahar ; Shiguang Shan. Springer Verlag, 2015. pp. 680-694 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).
@inproceedings{6d6ced40e4ac43cf80a4d8d2ca0ad6c6,
title = "Multi-target tracking with sparse group features and position using discrete-continuous optimization",
abstract = "Multi-target tracking of pedestrians is a challenging task due to uncertainty about targets, caused mainly by similarity between pedestrians, occlusion over a relatively long time and a cluttered background. A usual scheme for tackling multi-target tracking is to divide it into two sub-problems: data association and trajectory estimation. A reasonable approach is based on joint optimization of a discrete model for data association and a continuous model for trajectory estimation in a Markov Random Field framework. Nonetheless, usual solutions of the data association problem are based only on location information, while the visual information in the images is ignored. Visual features can be useful for associating detections with true targets more reliably, because the targets usually have discriminative features. In this work, we propose a combination of position and visual feature information in a discrete data association model. Moreover, we propose the use of group Lasso regularization in order to improve the identification of particular pedestrians, given that the discriminative regions are associated with particular visual blocks in the image. We find promising results for our approach in terms of precision and robustness when compared with a state-of-the-art method in standard datasets for multi-target pedestrian tracking.",
author = "Billy Peralta and Alvaro Soto",
year = "2015",
month = "1",
day = "1",
doi = "10.1007/978-3-319-16634-6_49",
language = "English",
isbn = "9783319166339",
series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",
publisher = "Springer Verlag",
pages = "680--694",
editor = "C.V. Jawahar and Shiguang Shan",
booktitle = "Computer Vision - 12th Asian Conference on Computer Vision, ACCV 2014, Revised Selected Papers",
address = "Germany",

}

Peralta, B & Soto, A 2015, Multi-target tracking with sparse group features and position using discrete-continuous optimization. En CV Jawahar & S Shan (eds.), Computer Vision - 12th Asian Conference on Computer Vision, ACCV 2014, Revised Selected Papers. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 9010, Springer Verlag, pp. 680-694, 12th Asian Conference on Computer Vision, ACCV 2014, Singapore, Singapore, 1/11/14. https://doi.org/10.1007/978-3-319-16634-6_49

Multi-target tracking with sparse group features and position using discrete-continuous optimization. / Peralta, Billy; Soto, Alvaro.

Computer Vision - 12th Asian Conference on Computer Vision, ACCV 2014, Revised Selected Papers. ed. / C.V. Jawahar; Shiguang Shan. Springer Verlag, 2015. p. 680-694 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 9010).

Resultado de la investigación: Conference contribution

TY - GEN

T1 - Multi-target tracking with sparse group features and position using discrete-continuous optimization

AU - Peralta, Billy

AU - Soto, Alvaro

PY - 2015/1/1

Y1 - 2015/1/1

N2 - Multi-target tracking of pedestrians is a challenging task due to uncertainty about targets, caused mainly by similarity between pedestrians, occlusion over a relatively long time and a cluttered background. A usual scheme for tackling multi-target tracking is to divide it into two sub-problems: data association and trajectory estimation. A reasonable approach is based on joint optimization of a discrete model for data association and a continuous model for trajectory estimation in a Markov Random Field framework. Nonetheless, usual solutions of the data association problem are based only on location information, while the visual information in the images is ignored. Visual features can be useful for associating detections with true targets more reliably, because the targets usually have discriminative features. In this work, we propose a combination of position and visual feature information in a discrete data association model. Moreover, we propose the use of group Lasso regularization in order to improve the identification of particular pedestrians, given that the discriminative regions are associated with particular visual blocks in the image. We find promising results for our approach in terms of precision and robustness when compared with a state-of-the-art method in standard datasets for multi-target pedestrian tracking.

AB - Multi-target tracking of pedestrians is a challenging task due to uncertainty about targets, caused mainly by similarity between pedestrians, occlusion over a relatively long time and a cluttered background. A usual scheme for tackling multi-target tracking is to divide it into two sub-problems: data association and trajectory estimation. A reasonable approach is based on joint optimization of a discrete model for data association and a continuous model for trajectory estimation in a Markov Random Field framework. Nonetheless, usual solutions of the data association problem are based only on location information, while the visual information in the images is ignored. Visual features can be useful for associating detections with true targets more reliably, because the targets usually have discriminative features. In this work, we propose a combination of position and visual feature information in a discrete data association model. Moreover, we propose the use of group Lasso regularization in order to improve the identification of particular pedestrians, given that the discriminative regions are associated with particular visual blocks in the image. We find promising results for our approach in terms of precision and robustness when compared with a state-of-the-art method in standard datasets for multi-target pedestrian tracking.

UR - http://www.scopus.com/inward/record.url?scp=84942511478&partnerID=8YFLogxK

U2 - 10.1007/978-3-319-16634-6_49

DO - 10.1007/978-3-319-16634-6_49

M3 - Conference contribution

SN - 9783319166339

T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

SP - 680

EP - 694

BT - Computer Vision - 12th Asian Conference on Computer Vision, ACCV 2014, Revised Selected Papers

A2 - Jawahar, C.V.

A2 - Shan, Shiguang

PB - Springer Verlag

ER -

Peralta B, Soto A. Multi-target tracking with sparse group features and position using discrete-continuous optimization. En Jawahar CV, Shan S, editores, Computer Vision - 12th Asian Conference on Computer Vision, ACCV 2014, Revised Selected Papers. Springer Verlag. 2015. p. 680-694. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). https://doi.org/10.1007/978-3-319-16634-6_49