Multiple camera based human motion estimation

Akira Utsumi, Hiroki Mori, Jun Ohya, Masahiko Yachida

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

Abstract

We propose a human motion detection method using multipleviewpoint images. We employ a simple elliptic model and a small number of reliable image features detected in multiple-viewpoint images to estimate the pose (position and normal axis) of a human body, where feature extraction is employed based on distance transformation. The COG (center of gravity) position and its distance value are extracted in the process. These features are robust against changes in human shapes caused by hand/leg bending and produce stable pose estimation results. After a pose estimation, a "best view" is selected based on the estimation result and further processing is performed including body-side detection and gesture recognition (in a 2D image of the selected view). This viewpoint selection approach can overcome the problem of self-occlusions. We confirmed the stability of the system through experiments.

Original languageEnglish
Title of host publicationLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
PublisherSpringer Verlag
Pages655-662
Number of pages8
Volume1352
ISBN (Print)3540639314, 9783540639312
Publication statusPublished - 1997
Externally publishedYes
Event3rd Asian Conference on Computer Vision, ACCV 1998 - Hong Kong, Hong Kong
Duration: 1998 Jan 81998 Jan 10

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume1352
ISSN (Print)03029743
ISSN (Electronic)16113349

Other

Other3rd Asian Conference on Computer Vision, ACCV 1998
CountryHong Kong
CityHong Kong
Period98/1/898/1/10

Fingerprint

Motion Estimation
Motion estimation
Camera
Cameras
Pose Estimation
Gesture recognition
Human Detection
Motion Detection
Gesture Recognition
Centre of gravity
Feature extraction
Gravitation
Occlusion
Feature Extraction
Processing
Human
Experiments
Estimate
Experiment
Model

ASJC Scopus subject areas

  • Computer Science(all)
  • Theoretical Computer Science

Cite this

Utsumi, A., Mori, H., Ohya, J., & Yachida, M. (1997). Multiple camera based human motion estimation. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) (Vol. 1352, pp. 655-662). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 1352). Springer Verlag.

Multiple camera based human motion estimation. / Utsumi, Akira; Mori, Hiroki; Ohya, Jun; Yachida, Masahiko.

Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). Vol. 1352 Springer Verlag, 1997. p. 655-662 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 1352).

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

Utsumi, A, Mori, H, Ohya, J & Yachida, M 1997, Multiple camera based human motion estimation. in Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). vol. 1352, Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 1352, Springer Verlag, pp. 655-662, 3rd Asian Conference on Computer Vision, ACCV 1998, Hong Kong, Hong Kong, 98/1/8.
Utsumi A, Mori H, Ohya J, Yachida M. Multiple camera based human motion estimation. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). Vol. 1352. Springer Verlag. 1997. p. 655-662. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).
Utsumi, Akira ; Mori, Hiroki ; Ohya, Jun ; Yachida, Masahiko. / Multiple camera based human motion estimation. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). Vol. 1352 Springer Verlag, 1997. pp. 655-662 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).
@inproceedings{56ac8d9d59424f2b87c9014edef66e92,
title = "Multiple camera based human motion estimation",
abstract = "We propose a human motion detection method using multipleviewpoint images. We employ a simple elliptic model and a small number of reliable image features detected in multiple-viewpoint images to estimate the pose (position and normal axis) of a human body, where feature extraction is employed based on distance transformation. The COG (center of gravity) position and its distance value are extracted in the process. These features are robust against changes in human shapes caused by hand/leg bending and produce stable pose estimation results. After a pose estimation, a {"}best view{"} is selected based on the estimation result and further processing is performed including body-side detection and gesture recognition (in a 2D image of the selected view). This viewpoint selection approach can overcome the problem of self-occlusions. We confirmed the stability of the system through experiments.",
author = "Akira Utsumi and Hiroki Mori and Jun Ohya and Masahiko Yachida",
year = "1997",
language = "English",
isbn = "3540639314",
volume = "1352",
series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",
publisher = "Springer Verlag",
pages = "655--662",
booktitle = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",

}

TY - GEN

T1 - Multiple camera based human motion estimation

AU - Utsumi, Akira

AU - Mori, Hiroki

AU - Ohya, Jun

AU - Yachida, Masahiko

PY - 1997

Y1 - 1997

N2 - We propose a human motion detection method using multipleviewpoint images. We employ a simple elliptic model and a small number of reliable image features detected in multiple-viewpoint images to estimate the pose (position and normal axis) of a human body, where feature extraction is employed based on distance transformation. The COG (center of gravity) position and its distance value are extracted in the process. These features are robust against changes in human shapes caused by hand/leg bending and produce stable pose estimation results. After a pose estimation, a "best view" is selected based on the estimation result and further processing is performed including body-side detection and gesture recognition (in a 2D image of the selected view). This viewpoint selection approach can overcome the problem of self-occlusions. We confirmed the stability of the system through experiments.

AB - We propose a human motion detection method using multipleviewpoint images. We employ a simple elliptic model and a small number of reliable image features detected in multiple-viewpoint images to estimate the pose (position and normal axis) of a human body, where feature extraction is employed based on distance transformation. The COG (center of gravity) position and its distance value are extracted in the process. These features are robust against changes in human shapes caused by hand/leg bending and produce stable pose estimation results. After a pose estimation, a "best view" is selected based on the estimation result and further processing is performed including body-side detection and gesture recognition (in a 2D image of the selected view). This viewpoint selection approach can overcome the problem of self-occlusions. We confirmed the stability of the system through experiments.

UR - http://www.scopus.com/inward/record.url?scp=84947568408&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84947568408&partnerID=8YFLogxK

M3 - Conference contribution

AN - SCOPUS:84947568408

SN - 3540639314

SN - 9783540639312

VL - 1352

T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

SP - 655

EP - 662

BT - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

PB - Springer Verlag

ER -