あいまいさを許容した自然物のモデル表現とその画像理解への応用

DOI Web Site Web Site 参考文献12件

書誌事項

タイトル別名
  • Modeling of Natural Objects Including Fuzziness and Application to Image Understanding
  • アイマイサ オ キョヨウシタ シゼンブツ ノ モデル ヒョウゲン ト ソノ ガ

この論文をさがす

抄録

We believe that an image understanding system should achieve results comparable to the human image understanding. Object recognition is an important part of image understanding. Typically, recognition is achieved by reasoning on the image processing results. In existing work, statistical, probabilistic and evidential (Dempster-Shafer theory) methods are used to treat the uncertainty associated with this reasoning. However, simplifying assumptions concerning the importance of, and the interaction between sources of evidence characterize these methods.<br>We propose a model based, object recognition method using fuzzy logic and capable of addressing the above issues. The object model is a part-of hierarchy. Each node is defined by fuzzy set valued attributes. When included, spatial relations between components are represented by fuzzy sets. A fuzzy measure captures interactions between attributes. Image processing results are propagated to a node by integration with respect to this fuzzy measure. When spatial relations are included, the recognition of nodes and relations are propagated to the level of a triplet (two nodes and a spatial relation). Integration with respect to a fuzzy measure on the collection of items (triplets, or nodes) propagates the recognition to higher levels. The fuzzy framework offers a unified approach to representation and reasoning while incorporating the user's subjectivity. Results of initial experiments are discussed.

収録刊行物

参考文献 (12)*注記

もっと見る

キーワード

詳細情報 詳細情報について

問題の指摘

ページトップへ