Volume 14 Number 10 (Oct. 2019)
Home > Archive > 2019 > Volume 14 Number 10 (Oct. 2019) >
JSW 2019 Vol 14(10): 437-448 ISSN: 1796-217X
doi: 10.17706/jsw.14.10.437-448

Fast 3D Post Estimation of Human Based on Optical Flow and Particle Filter

Jiaying Lan, Guoheng Huang*, Lianglun Cheng

School of Computer Science, Guangdong University of Technology,Guangzhou, Guangdong, China

Abstract— Aiming at the high computational complexity of the 3D pose estimation algorithm in the deep learning field, we propose a fast human poseestimation algorithm based on optical flow and particle filter. The temporal correlation between video frames is applied to the algorithm. The first frame of the video is defined as a keyframe, which will serve as the output of the 3D pose estimate. Then the next frame is determined by the key frame algorithm whether it is a key frame. The key frame is estimated by the 3D human pose estimation algorithm, and theoutput result of the key frame is propagated to the non-key frame through the optical flow mechanism.Non-key frames are subjected to pose estimation through particle filter. In the 3D human pose estimation problem, we propose a unified equation for 3D human pose estimation from the RGB image, combining 2D joint estimation and 3D posereconstruction. Theproposed approach outperforms all state-of-the-art methods on Human3.6m achieving a relative error reduction greater than 30% on average. Our method significantly improves detection performance compared to the original algorithm, and the detection speed can be increased by an average of 43.75%

Index Terms—Dpose estimation, optical flow, multi-subspaceparticlefilter.


Cite: Jiaying Lan, Guoheng Huang, Lianglun Cheng, "Fast 3D Post Estimation of Human Based on Optical Flow and Particle Filter," Journal of Software vol. 14, no. 10, pp. 437-448, 2019.

General Information

ISSN: 1796-217X (Online)
Frequency:  Quarterly
Editor-in-Chief: Prof. Antanas Verikas
Executive Editor: Ms. Yoyo Y. Zhou
Abstracting/ Indexing: DBLP, EBSCO, CNKIGoogle Scholar, ProQuest, INSPEC(IET), ULRICH's Periodicals Directory, WorldCat, etc
E-mail: jsweditorialoffice@gmail.com
  • Mar 01, 2024 News!

    Vol 19, No 1 has been published with online version    [Click]

  • Jan 04, 2024 News!

    JSW will adopt Article-by-Article Work Flow

  • Apr 01, 2024 News!

    Vol 14, No 4- Vol 14, No 12 has been indexed by IET-(Inspec)     [Click]

  • Apr 01, 2024 News!

    Papers published in JSW Vol 18, No 1- Vol 18, No 6 have been indexed by DBLP   [Click]

  • Nov 02, 2023 News!

    Vol 18, No 4 has been published with online version   [Click]