In recent years, the number of the occurrence of malicious crimes, such as the use of identity theft to withdraw cash from ATMs has significantly increased. To prevent these crimes, the construction of security systems is indispensable. However, most of security systems currently employed have several problems to be solved. In the security system using biometric personal authentification, the placement of hand should be posed correctly with right orientation. Therefore, in this study, we solved this problem by constructing two types of systems that carry out personal authentification using the image information from hand that is placed in various positions and orientations. The first system extracts palm print from images of hand placed in various positions and orientations and performs personal authentification using template matching. The second system extracts as feature amount, such as the length and thickness of finger and performs personal authentification with support vector machine. Using 240 images obtained from the hands of 48 registered subjects, we carried out personal authentification utilizing both systems. The results show that both systems led to a reduction in the false rejection rate to a few percent with the false acceptance rate of 0%. These results confirm the usability of both systems.

114 工学部 (旧) ・工学研究科 ＞ 学位論文 ＞ 修士論文