Globally Optimal Object Tracking with Complementary Use of Single Shot Multibox Detector and Fully Convolutional Network

Jinho Lee, Brian Kenji Iwana, Shouta Ide, Hideaki Hayashi, Seiichi Uchida

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

Object tracking is one of the most important but still difficult tasks in computer vision and pattern recognition. The main difficulties in the tracking task are appearance variation of target objects and occlusion. To deal with those difficulties, we propose a object tracking method combining Single Shot Multibox Detector (SSD), Fully Convolutional Network (FCN) and Dynamic Programming (DP). SSD and FCN provide a probability value of the target object which allows for appearance variation within each category. DP provides a globally optimal tracking path even with severe occlusions. Through several experiments, we confirmed that their combination realized a robust object tracking method. Also, in contrast to traditional trackers, initial position and a template of the target do not need to be specified. We show that the proposed method has a higher performance than the traditional trackers in tracking various single objects through video frames.

Original languageEnglish
Title of host publicationImage and Video Technology - 8th Pacific-Rim Symposium, PSIVT 2017, Revised Selected Papers
EditorsCarlos Hitoshi, Manoranjan Paul, Qingming Huang
PublisherSpringer Verlag
Pages110-122
Number of pages13
ISBN (Print)9783319757858
DOIs
Publication statusPublished - Jan 1 2018
Event8th Pacific-Rim Symposium on Image and Video Technology, PSIVT 2017 - Wuhan, China
Duration: Nov 20 2017Nov 24 2017

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume10749 LNCS
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Other

Other8th Pacific-Rim Symposium on Image and Video Technology, PSIVT 2017
CountryChina
CityWuhan
Period11/20/1711/24/17

    Fingerprint

All Science Journal Classification (ASJC) codes

  • Theoretical Computer Science
  • Computer Science(all)

Cite this

Lee, J., Iwana, B. K., Ide, S., Hayashi, H., & Uchida, S. (2018). Globally Optimal Object Tracking with Complementary Use of Single Shot Multibox Detector and Fully Convolutional Network. In C. Hitoshi, M. Paul, & Q. Huang (Eds.), Image and Video Technology - 8th Pacific-Rim Symposium, PSIVT 2017, Revised Selected Papers (pp. 110-122). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 10749 LNCS). Springer Verlag. https://doi.org/10.1007/978-3-319-75786-5_10