대한전기학회:학술대회논문집 (Proceedings of the KIEE Conference)
- 대한전기학회 1988년도 전기.전자공학 학술대회 논문집
- /
- Pages.683-686
- /
- 1988
Recognition of Profile Contours of Human Face by Approximation - Recognition
초록
In the recognition of similar patterns like profile contours of human faces, feature measure plays important role. We extracted effective and general feature by B-spline approximation. The nodes and vertices of the approximated curve are normalized and used as features. Since the features have both local property of curvature extrema and global property by B-spline approximation, they are superior to those of curvature extrema of the profile contour. For the image data of six sets of 56 persons, some of which are ill-made, averaged accuracy rate of 97.6 % is obtained in recognizing combinational 333 test samples.
키워드