AN OBJECT DISTINCTION METHOD BY CLUSTERING OF MOTION CHARACTERISTICS USING SIFT TRACKING

摘要：

In this paper, an object distinguish method is proposed. This method is based on clustering of motion characteristics which are obtained by tracking SIFT feature points. SIlT (Scale Invariant Feature Transform) is a feature detection method of an image. This method finds features as local maximum point of DoG ( Differ ence of Gaussian) image. DoG is generated as layered difference images of Gaussian filtered images on differ ent scales. The motion characteristics are described as a vector of transition lengths in the image. The experi mental results show that the proposed method can distinguish objects which have different motion characteris tics.

关键词： Scene Recognition Object Distinction Feature Tracking SIFT

作者: Hitoshi Yamauchi Akihiro Kanagawa

作者单位: Okayama Prefectural University, Okayama 719 - 1197, Japan

会议类型: 国际会议

会议名称: The Ninth International Conference on Industrial Management(第九届工业管理国际会议 ICIM2008)

会议地点: 日本大阪

会议语种:英文

页码: 867-871

在线出版日期: 2008-09-16（万方平台首次上网日期，不代表论文的发表时间）