Pattern Recognition of Human Faces and Seals

• Main Authors: Jyun-Neng Liao, 廖俊能
• Other Authors: Chien-Hsing Lee
• Format: Others
• Language: zh-TW
• Published: 2008
• Online Access: http://ndltd.ncl.edu.tw/handle/42363284470611982200

碩士 === 國立成功大學 === 系統及船舶機電工程學系碩博士班 === 96 === A 2D image captured by a CCD is generally composed by pixels, e.g., pattern recognition of seals and human faces. The seal has firm external contour with classifying by its internal words; the human face has inextricable complexity of external contour with an internal facial feature. This thesis will investigate the difference of pattern recognition of seals and human faces. In this thesis, many feature extraction and classification methods are used to analyze the 2D image with a height obtained by the photometric stereo method (called 2.5D face) and the 2D image with pixels directly captured by a CCD (called 2D face). One of the feature extraction methods, namely Delaunay triangulation, uses the external contour and internal words of a seal to construct the structure of the triangular network without affecting by the capturing position and angle. Moreover, the Delaunay triangulation for face regnition uses the external contour and the height of 2.5D face to increase or decrease its characteristics without affecting by the light and facial expression. Finally, this thesis has developed a user interface of pattern recognition based on the images of 2.5D human faces and seals.