品質工学
Online ISSN : 2189-9320
Print ISSN : 2189-633X
ISSN-L : 2189-633X
事例研究
MTシステムを用いた顔画像による個人識別の研究(2)
-マルチMT法による個人識別-
野田 健次田村 希志臣斎藤 之男矢野 宏
著者情報
ジャーナル フリー

2011 年 19 巻 1 号 p. 53-58

詳細
抄録

There is currently a market demand for systems that can identify faces objectively. Research on facial identification using coordinates of features such as the eyes and nose has made personal identification possible. However, the feature points at which the coordinates are taken have been selected subjectively,and only a small part of the image information has been used. In this study,subjective selection of feature points was avoided to the maximum extent possible by using the facial image pixels themselves as feature points. Since the number of such feature points is very large, the multilevel Mahalanobis-Taguchi (MT) system was used to define a unit space. The results confirmed the feasibility of personal identification based on pixel information. It is anticipated that pixel-based image recognition techniques will be applicable not only to facial image recognition, but also to more general-purpose image recognition.

著者関連情報
© 2011 一般社団法人 品質工学会
前の記事 次の記事
feedback
Top