英文摘要 | Fingerprint recognition technology is one of the far-ranging application technologies in the biometrics, and has long history. In the past, it was mainly used in the criminal act and a cout of justice and few people known this techology. With the development of computer science, fingerprint recognition has extensive outlook servicing people. An automatic fingerprint identification system (AFIS) is composed of on-line and off-line parts. In the off-line part, the sensor capture the fingerprint image, the feature information is extracted, at last storage the features into database which are called as fingerprint template. In the on-line part, the sensor read the fingerprint image, and features are detected as in off-line part, and the features are matched with those in the template, judge whether the input features and template features are coming frome same fingerprint of same person. Some poor quality images exist in the captured images, and these poor quality images need to be ehnaced. After enhancing the image, mainly extract and select the feature. Feature selection is essence of influenceing the recognition performance. In an AFIS, there are two ways to select the features, one is the minutiae model proposed by FBI, including those system which use other information beside minutiae. The other is based on the texture of the fingerprint, which use the texture features to recognize the fingerprint image. After selecting suitable features, a fingerprint matching algorithm is built up based on the features. There are many algorithms which are applicable in the practice in field of fingerprint verification, whether fingerprint classification, fingerprint enhancement, and fingerprint matching, but still some problems remain to be valuable for research. This paper studies the fingerprint image enhancement, feature selection, recognition performance and minutiae matching.The main work of this dissertation is as follows: (1) We propose a nonlinear diffusion filter to enhance fingerprint image and extract the minutiae. The algorithm uses the structure tensor to construct diffusion tensor, and make the filter diffuse along the ridge direction. In order to improve the computing efficiency, according to the discrete format of the diffusion tensor in a window size of (2n+l)~ (2n+l), split the diffusion filter into 4n directions, after that, in each direction compute the fingerprint enhancement image with Thomas-Gaussian elimination algorithm. The advantage of this algorithm is to connect the breaking ridges in an effective way, and cannot change the continuity and singularity of the ridge.This method can enhance the image captured in off-line part in AFIS. (2)We propose dyadic scale space (DSS) theory in the computer science to enhance the on-line fingerprint image.DSS algorithm decomposes the fingerprint image into a series of images and organize th |
修改评论