Multiple camera based human motion estimation

Akira Utsumi, Hiroki Mori, Jun Ohya, Masahiko Yachida

研究成果

抄録

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.

本文言語English
ホスト出版物のタイトルComputer Vision - ACCV 1998 - 3rd Asian Conference on Computer Vision, Proceedings
編集者Roland Chin, Ting-Chuen Pong
出版社Springer Verlag
ページ655-662
ページ数8
ISBN(印刷版)3540639314, 9783540639312
DOI
出版ステータスPublished - 1997 1月 1
外部発表はい
イベント3rd Asian Conference on Computer Vision, ACCV 1998 - Hong Kong, Hong Kong
継続期間: 1998 1月 81998 1月 10

出版物シリーズ

名前Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
1352
ISSN(印刷版)0302-9743
ISSN(電子版)1611-3349

Other

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

ASJC Scopus subject areas

  • 理論的コンピュータサイエンス
  • コンピュータ サイエンス(全般)

フィンガープリント

「Multiple camera based human motion estimation」の研究トピックを掘り下げます。これらがまとまってユニークなフィンガープリントを構成します。

引用スタイル