Research On The Key Technologies For Automatic Fingerprint Identification System

Because human fingerprint is permanent and unique, fingerprint verification has become one of the most widely applied techniques in biometric verification technology and the technique of identifying people through fingerprint minutiae is becoming more and more popular. Fingerprint recognition has been extensively studied and many advances have been made on it. However due to the complexity and uncertainty of fingerprint identification, there are still many problems to be resolved for further improvement of Automatic Fingerprint Identification System. In this paper, fingerprint image quality evaluation algorithms, fingerprint image preprocessing algorithms, fingerprint matching algorithm are analyzed and discussed in detail. The main contributions of the dissertation are as following:1. The dissertation investigated fingerprint image quality evaluation and proposed a new method to evaluate fingerprint image quality on the basis of discriminant factors. The order of discriminant factors is in accordance with the impact on the final evaluation results. If the fingerprint image doesn't meet the requirement, the process of evaluation is end. The user is requested to input other one again and the suggestion for revision is proposed at the same time.2. Fingerprint image preprocessing algorithm is studied. The types of fingerprint image regions are analyzed and multistage segmentation is proposed. Foreground and background are separated by primary segmentation. Secondary segmentation which uses linear support vector machine follows the steps of primary segmentation to cut the unrecoverable regions from the foreground of primary segmentation, and thus is able to avoid extraction of false feature.3. Fingerprint feature matching algorithm is studied, and a triangle-based fingerprint minutia matching algorithm is proposed. The whole matching process is divided into three steps. First, in terms of the specified method, find two congruent triangles within error tolerance from the entire extracted fingerprint minutia and the fingerprint minutiae templates to be matched, with each singular point as one vertex of a triangle. Then according to the relative positions of two triangles, translate and rotate the triangle formed by the minutiae extracted till the two triangles overlap completely within error tolerance and determine the translation and rotation parameter with which all the fingerprint minutiae extracted are translated and rotated. Finally, elastic matching algorithm is adopted to compute global feature distance for minutiae's corresponding relation and matching score determination.