人物の流れに着目した大局的な行動パターンの逐次学習・認識

カテゴリ: 論文誌(論文単位)

グループ名: 【D】産業応用部門

発行日: 2011/06/01

タイトル(英語): Sequential Learning and Recognition of Comprehensive Behavioral Patterns Based on Flow of People

著者名: 宜保　達哉（大阪府立大学大学院工学研究科），青木　茂樹（大阪府立大学総合教育研究機構），宮本　貴朗（大阪府立大学総合教育研究機構），岩田　基（大阪府立大学大学院工学研究科），汐崎　陽（大阪府立大学大学院工学研究科）

著者名(英語): Tatsuya Gibo (Graduate School of Engineering, Osaka Prefecture University), Shigeki Aoki (Faculty of Liberal Arts and Sciences, Osaka Prefecture University), Takao Miyamoto (Faculty of Liberal Arts and Sciences, Osaka Prefecture University), Motoi Iwata (Graduate School of Engineering, Osaka Prefecture University), Akira Shiozaki (Graduate School of Engineering, Osaka Prefecture University)

キーワード: 行動パターン認識，オプティカルフロー，動的計画法　　behavioral pattern recognition，optical flow，dynamic programming

要約(英語): Recently, surveillance cameras have been set up everywhere, for example, in streets and public places, in order to detect irregular situations. In the existing surveillance systems, as only a handful of surveillance agents watch a large number of images acquired from surveillance cameras, there is a possibility that they may miss important scenes such as accidents or abnormal incidents. Therefore, we propose a method for sequential learning and the recognition of comprehensive behavioral patterns in crowded places. First, we comprehensively extract a flow of people from input images by using optical flow. Second, we extract behavioral patterns on the basis of change-point detection of the flow of people. Finally, in order to recognize an observed behavioral pattern, we draw a comparison between the behavioral pattern and previous behavioral patterns in the database. We verify the effectiveness of our approach by placing a surveillance camera on a campus.

本誌: 電気学会論文誌D（産業応用部門誌） Vol.131 No.6 （2011）

本誌掲載ページ: 820-828 p

原稿種別: 論文／日本語

電子版へのリンク: https://www.jstage.jst.go.jp/article/ieejias/131/6/131_6_820/_article/-char/ja/