Spatial pixels selection and inter-frame combined likelihood based observation for 60 fps 3d tracking of twelve volleyball players on gpu

Yiming Zhao, Xina Cheng, Takeshi Ikenaga

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

2 Citations (Scopus)

Abstract

3D players tracking plays an important role in sports analysis. Tracking of players contributes to high level game analysis such as tactic analysis and commercial applications such as TV contents. Many services like sports live and broadcasting have strict limitation on processing time, thus real-time implementation for 3D players tracking is necessary. This paper proposes a particle filter based 60 fps multi-view volleyball players tracking system on GPU platform. There are three proposals: body region constraint prediction, spatial pixels selection and inter-frame combined likelihood. The body region constraint prediction uses player’s body region as limitation in prediction to increase tracking accuracy. The spatial pixels selection method selects pixels for likelihood calculating to reduce calculation amount in spatial space. The inter-frame observation method does particle filter algorithm with two frames each time to reduce calculation amount in temporal space. Our experiments are based on videos of the Final and Semi-Final Game of 2014 Japan Inter High School Games of Men’s Volleyball in Tokyo Metropolitan Gymnasium. On the GPU device GeForce GTX 1080Ti, our tracking system achieves real-time on 60 fps videos and keeps the tracking accuracy higher than 97%.

Original languageEnglish
Title of host publicationAdvances in Multimedia Information Processing – PCM 2018 - 19th Pacific-Rim Conference on Multimedia, 2018, Proceedings
EditorsWen-Huang Cheng, Toshihiko Yamasaki, Chong-Wah Ngo, Richang Hong, Meng Wang
PublisherSpringer-Verlag
Pages716-726
Number of pages11
ISBN (Print)9783030007669
DOIs
Publication statusPublished - 2018 Jan 1
Event19th Pacific-Rim Conference on Multimedia, PCM 2018 - Hefei, China
Duration: 2018 Sep 212018 Sep 22

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume11165 LNCS
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Other

Other19th Pacific-Rim Conference on Multimedia, PCM 2018
CountryChina
CityHefei
Period18/9/2118/9/22

    Fingerprint

Keywords

  • 3D volleyball players tracking
  • GPU acceleration
  • Particle filter
  • Sports analysis

ASJC Scopus subject areas

  • Theoretical Computer Science
  • Computer Science(all)

Cite this

Zhao, Y., Cheng, X., & Ikenaga, T. (2018). Spatial pixels selection and inter-frame combined likelihood based observation for 60 fps 3d tracking of twelve volleyball players on gpu. In W-H. Cheng, T. Yamasaki, C-W. Ngo, R. Hong, & M. Wang (Eds.), Advances in Multimedia Information Processing – PCM 2018 - 19th Pacific-Rim Conference on Multimedia, 2018, Proceedings (pp. 716-726). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 11165 LNCS). Springer-Verlag. https://doi.org/10.1007/978-3-030-00767-6_66