Research And Realization Of Fingerprint Identification Algorithm

In recent years, Automatic identification technique has been an emerging approach in real life. Especially, biometric authentication has drawn much attention. Due to its strong reliability, many modern approaches have been proposed by researchers and many biometric-related products have been developed.Fingerprint identification method has been researched and applied widely as a major technique in biometric domain. But the factors limiting the performance of fingerprint verification system mainly include fingerprint image quality, reliability of fingerprint minutiae, and the performance of fingerprint matching methods. Due to the uncontrollability of fingerprint image quality, the performance of whole fingerprint authentication system is mainly determined by the minutiae extraction algorithm and the fingerprint matching methods. Although fingerprint matching methods have been explored deeply, current matching algorithms can not still satisfy the real need completely. Most of current fingerprint matching schemes are based on minutiae. It is a well-known point pattern matching problem. This paper focuses on the problem and proposes a novel matching method. We implement them and demonstrate their effectiveness. The main innovative contributions to fingerprint matching in our paper are as follows:1) Proposing a fingerprint image quality evaluation algorithm. This algorithm is simple and efficient, without calculating the orientation field of fingerprint image. In accordance with the statistical distribution of gray value of fingerprint image, a threshold will be chosen based on the characteristics of different fingerprints, then low quality fingerprint images will be kept out. The experimental results show that it's a good quality assessment algorithm, and has good usability.2) Researching some fingerprint image preprocessing techniques, such as normalizing, orientation field, enhancing, thinning, post-thinning and so on, and some schemes are proposed. Orientation field is obtained by using isotropy Sobel operator, and the enhancement of fingerprint image is based on Gabor filters which have both frequency-selective and orientation-selective properties, we improve the fast-thinning algorithm based on mathematic morphologic to thin fingerprint image, and delete short line and burr in post-thinning step which may bring pseudo minutiae.3) Proposing a novel minutia local structure which makes use of neighboring minutiae and global orientation fields. A new fingerprint matching technique based on our proposed structure is given. Because the matching algorithm uses more rich information including neighboring minutiae and orientation field, it gets better experimental results than some matching methods that only uses minutiae information or orientation information.