The Research Of Planar Characteristics In Automatic Fingerprint Verification Algorithm

In current society, the trend of digital and recessive personal identity is more and more distinct. How to automatically identify a person with high level of confidence to ensure the information security has been a critical social problem. Modem personal identification technology must not only take a very high security, full automatic identification, simple management configuration and easy to use are also its indispensable features. Biometrics, which refers to automatic identification of a person based on his physiological or behavioral characteristics, is the most suitable candidate. Automatic fingerprint identification is one of the most reliable and widely used biometric technology. Most popular fingerprint verification algorithms focus on the local minutiae such as ridge endings and ridge bifurcations. They match different minutiae templates to verify specified fingerprints. However, because local minutiae are vulnerable to noises, lots of spurious minutiae will be generated when processing the low-quality images, this leads errors to fingerprint match process. It's almost impossible to solve the problem only by fingerprint post-process. In order to make the fingerprint verification algorithm more robust to the low-quality images, we try to make use of the ridgelines as the planar characteristics. The planar characteristics are global characteristics in the fingerprint images and are less vulnerable to the local noise. We believe that by making use of the planar characteristics, some of the errors will be eliminated, resulting in an improvement in algorithm performance.In this thesis, we focus on the retrieval , description and the similarity of the planar characteristics. We have identified and explored the following issues: (i) image preprocess including image normalization, image' segmentation and image enhancement which transform the raw fingerprint images to skeleton; (ii) image post-processing which fixes the damaged ridge structures and eliminates false minutiae; (iii) the retrieval and description of the planar characteristics of fingerprint images which get the planar characteristics from skeleton; (/v) exploration of the similarity between different planar characteristics of fingerprint images. Finally we apply the planar characteristics to the fingerprint matching process to replace the curve segments used by traditional minutiae-based algorithm. The experiment results are fairly well.