Object tracking with multiple instance learning and gaussian mixture model

Na Li, Xiangmo Zhao, Daxiang Li, Jing Wang, Bendu Bai

Research output: Contribution to journalArticle

1 Citation (Scopus)

Abstract

Recently, Multiple Instance Learning (MIL) technique has been introduced for object tracking applications, which has shown its good performance to handle drifting problem. While some instances in positive bags not only contain objects, but also contain the background, it is not reliable to simply assume that each feature of instances in positive bags obeys a single Gaussian distribution. In this paper, a tracker based on online multiple instance boosting has been developed, which employs Gaussian Mixture Model (GMM) and single Gaussian distribution respectively to model features of instances in positive and negative bags. The differences between samples and the model are integrated into the process of updating the parameters for GMM. With the Haar-like features extracted from the bags, a set of weak classifiers are trained to construct a strong classifier, which is used to track the object location at a new frame. And the classifier can be updated online frame by frame. Experimental results have shown that our tracker is more stable and efficient when dealing with the illumination, rotation, pose and appearance changes.

LanguageEnglish
Pages4465-4477
Number of pages13
JournalJournal of Information and Computational Science
Volume12
Issue number11
DOIs
Publication statusPublished - 20 Jul 2015

Fingerprint

Classifiers
Gaussian distribution
learning
Lighting
performance

Cite this

Li, Na ; Zhao, Xiangmo ; Li, Daxiang ; Wang, Jing ; Bai, Bendu. / Object tracking with multiple instance learning and gaussian mixture model. In: Journal of Information and Computational Science. 2015 ; Vol. 12, No. 11. pp. 4465-4477.
@article{3e45d3470e864464a62f9af4db2420d2,
title = "Object tracking with multiple instance learning and gaussian mixture model",
abstract = "Recently, Multiple Instance Learning (MIL) technique has been introduced for object tracking applications, which has shown its good performance to handle drifting problem. While some instances in positive bags not only contain objects, but also contain the background, it is not reliable to simply assume that each feature of instances in positive bags obeys a single Gaussian distribution. In this paper, a tracker based on online multiple instance boosting has been developed, which employs Gaussian Mixture Model (GMM) and single Gaussian distribution respectively to model features of instances in positive and negative bags. The differences between samples and the model are integrated into the process of updating the parameters for GMM. With the Haar-like features extracted from the bags, a set of weak classifiers are trained to construct a strong classifier, which is used to track the object location at a new frame. And the classifier can be updated online frame by frame. Experimental results have shown that our tracker is more stable and efficient when dealing with the illumination, rotation, pose and appearance changes.",
keywords = "Gaussian mixture model, Multiple instance learning, Object tracking",
author = "Na Li and Xiangmo Zhao and Daxiang Li and Jing Wang and Bendu Bai",
year = "2015",
month = "7",
day = "20",
doi = "10.12733/jics20106217",
language = "English",
volume = "12",
pages = "4465--4477",
journal = "Journal of Information and Computational Science",
issn = "1548-7741",
publisher = "Binary Information Press",
number = "11",

}

Object tracking with multiple instance learning and gaussian mixture model. / Li, Na; Zhao, Xiangmo; Li, Daxiang; Wang, Jing; Bai, Bendu.

In: Journal of Information and Computational Science, Vol. 12, No. 11, 20.07.2015, p. 4465-4477.

Research output: Contribution to journalArticle

TY - JOUR

T1 - Object tracking with multiple instance learning and gaussian mixture model

AU - Li, Na

AU - Zhao, Xiangmo

AU - Li, Daxiang

AU - Wang, Jing

AU - Bai, Bendu

PY - 2015/7/20

Y1 - 2015/7/20

N2 - Recently, Multiple Instance Learning (MIL) technique has been introduced for object tracking applications, which has shown its good performance to handle drifting problem. While some instances in positive bags not only contain objects, but also contain the background, it is not reliable to simply assume that each feature of instances in positive bags obeys a single Gaussian distribution. In this paper, a tracker based on online multiple instance boosting has been developed, which employs Gaussian Mixture Model (GMM) and single Gaussian distribution respectively to model features of instances in positive and negative bags. The differences between samples and the model are integrated into the process of updating the parameters for GMM. With the Haar-like features extracted from the bags, a set of weak classifiers are trained to construct a strong classifier, which is used to track the object location at a new frame. And the classifier can be updated online frame by frame. Experimental results have shown that our tracker is more stable and efficient when dealing with the illumination, rotation, pose and appearance changes.

AB - Recently, Multiple Instance Learning (MIL) technique has been introduced for object tracking applications, which has shown its good performance to handle drifting problem. While some instances in positive bags not only contain objects, but also contain the background, it is not reliable to simply assume that each feature of instances in positive bags obeys a single Gaussian distribution. In this paper, a tracker based on online multiple instance boosting has been developed, which employs Gaussian Mixture Model (GMM) and single Gaussian distribution respectively to model features of instances in positive and negative bags. The differences between samples and the model are integrated into the process of updating the parameters for GMM. With the Haar-like features extracted from the bags, a set of weak classifiers are trained to construct a strong classifier, which is used to track the object location at a new frame. And the classifier can be updated online frame by frame. Experimental results have shown that our tracker is more stable and efficient when dealing with the illumination, rotation, pose and appearance changes.

KW - Gaussian mixture model

KW - Multiple instance learning

KW - Object tracking

UR - http://www.scopus.com/inward/record.url?scp=84939447094&partnerID=8YFLogxK

U2 - 10.12733/jics20106217

DO - 10.12733/jics20106217

M3 - Article

VL - 12

SP - 4465

EP - 4477

JO - Journal of Information and Computational Science

T2 - Journal of Information and Computational Science

JF - Journal of Information and Computational Science

SN - 1548-7741

IS - 11

ER -