Online tracking based on multiple appearances model

Shuo Tang, Long Fei Zhang, Jia Li Yan, Xiang Wei Tan, Gang Yi Ding

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

Abstract

Tracking target in a long-term is still a big challenge in computer vision. In recent research, many researchers pay much attention on updating current appearance of tracking target to build one online appearance model. However, one appearance model is always not enough to describe historical appearance information especially for long-term tracking task. In this paper, we propose an online multiple appearances model based on Dirichlet Process Mixture Model (DPMM), which can make different appearance representations of the tracking target grouped dynamically and in an unsupervised way. Since DPMM's appealing properties are characterized by Gibbs sampling and Gibbs sampling costs too much, we proposed an online Bayesian learning algorithm instead of Gibbs sampling to reliably and efficiently learn a DPMM from scratch through sequential approximation in a streaming fashion to adapt new tracking targets. Experiments on multiple challenging benchmark public dataset demonstrate the proposed tracking algorithm performs favorably against the state-of-the-art.

Original languageEnglish
Title of host publicationProceedings - 2016 International Conference on Information System and Artificial Intelligence, ISAI 2016
PublisherInstitute of Electrical and Electronics Engineers Inc.
Pages634-637
Number of pages4
ISBN (Electronic)9781509015856
DOIs
Publication statusPublished - 12 Jan 2017
Event2016 International Conference on Information System and Artificial Intelligence, ISAI 2016 - Hong Kong, China
Duration: 24 Jun 201626 Jun 2016

Publication series

NameProceedings - 2016 International Conference on Information System and Artificial Intelligence, ISAI 2016

Conference

Conference2016 International Conference on Information System and Artificial Intelligence, ISAI 2016
Country/TerritoryChina
CityHong Kong
Period24/06/1626/06/16

Keywords

  • Multiple appearance model
  • Object tracking
  • Online Dirichlet process mixture model

Fingerprint

Dive into the research topics of 'Online tracking based on multiple appearances model'. Together they form a unique fingerprint.

Cite this