| Summary: | 碩士 === 國立中山大學 === 電機工程學系研究所 === 102 === Biometric features are innate characteristics of human beings, including fingerprint, palm-print, voice-print, face, iris and retina. They are unique, difficult to counterfeit, and widely used in the fields of person identification, crime investigation, fortune telling, data encryption, computer login and facility access. With the rapid development of technologies, more delicate authorization mechanism is required to build reliable information systems and avoid potential confidentiality disclosure. Therefore, user identification capability is indispensable for designing modern and future information systems. In our daily life, keys, IC cards and RFID cards are usually accompanied by the risk of missing and embezzling. Fingerprint is one of the most widely used biometric features for the immigration offices, business facilities and personal properties around the world. By the use of local binary pattern, it is our sincere hope to design a trustworthy fingerprint recognition system to enhance data integrity.
In this thesis, two input devices, a scanner and a mobile camera, are used to acquire the fingerprint images. For the scanner scenario, three levels of weights with light, moderate and heavy pressures are applied to rub the inked fingerprints on the paper, and then the Epson Perfection V33 scanner is utilized to capture the images. For the mobile camera approach, the Sony Xperia ZL C6502 mobile phone with Sony Exmor RS™ camera is used to obtain the fingerprint images. Image binarization, noise reduction and boundary selection are applied in the preprocessing steps, and local binary pattern is then adopted to extract the texture features. Under the Intel Core 2 Duo P8700 @ 2.53GHz personal computer and Windows7 operating system environment, correct rates of 100% and 100% can be obtained using the scanner and mobile camera schemes respectively for a database with 11 laboratory users. In addition, the system is tested on the Cross Match Technology database of 51 users, Digital Persona database of 65 users and AuthenTec databases of 21 users and 16 users, the correct fingerprint recognition rates of 85.62%, 78.46%, 95.24% and 100.0% can be achieved respectively.
|
|---|
