Level Set Based Online Visual Tracking via Convolutional Neural Network

Xiaodong Ning, Lixiong Liu*

*Corresponding author for this work

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

1 Citation (Scopus)

Abstract

In this paper, we propose a level set tracking algorithm, which integrates the information of the original frame and the confidence predicted by the deep feature based detector. First, we extract features from convolutional neural network and select part of them to avoid redundancy. Secondly, the features are used to generate a confidence map of the tracked object through the detector. And then the confidence along with the original frame is applied in level set model to acquire the segmentation result. We introduce an outlier rejection scheme to further improve the result. Finally, updating is employed to the detector to adapt to the changes in the video. One important contribution of our work is to use the deep features in confidence prediction, particularly the usage of low-level features in the neural network. Experimental results show that our model delivers a better performance than the state-of-the-art on a series of challenging videos.

Original languageEnglish
Title of host publicationNeural Information Processing - 24th International Conference, ICONIP 2017, Proceedings
EditorsDerong Liu, Shengli Xie, El-Sayed M. El-Alfy, Dongbin Zhao, Yuanqing Li
PublisherSpringer Verlag
Pages280-290
Number of pages11
ISBN (Print)9783319700892
DOIs
Publication statusPublished - 2017
Event24th International Conference on Neural Information Processing, ICONIP 2017 - Guangzhou, China
Duration: 14 Nov 201718 Nov 2017

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume10636 LNCS
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Conference

Conference24th International Conference on Neural Information Processing, ICONIP 2017
Country/TerritoryChina
CityGuangzhou
Period14/11/1718/11/17

Keywords

  • Convolutional neural network
  • Deep feature
  • Level set
  • Object tracking

Fingerprint

Dive into the research topics of 'Level Set Based Online Visual Tracking via Convolutional Neural Network'. Together they form a unique fingerprint.

Cite this