Algorithm Research Of Fingerprint Recognition System

Biometric identification as an effective identification technology quickly gets the public accepted. A biometric system is essentially a pattern recognition system that operates by acquiring biometric data from an individual, extracting a feature set from the acquired data and comparing this feature set against the template set in the database to see if they are from the same individual. Fingerprint recognition is the oldest and most widely used in the application of biometric recognition, because each person has a unique set of fingerprints, fingerprints become more and more common for personal identification, Fingerprint recognition are widely used in many important occasions because of the own uniqueness, stability and convenience of collection and other characteristics. Since the computer has been widely used, Fingerprint identification has been a problem in the field of computer science research. Automatic Fingerprint Identification System typically consists of image acquisition, pretreatment, feature extraction and matching, etc. Now the research on fingerprint identification technology mostly focus on the follow several aspects involving the validity of the fingerprint recognition system, recognition accuracy and speed. Among them, the fingerprint image pre-processing and feature matching algorithm are the two processes of high complexity, the most time-consuming, so currently the most studieds are focused on the two parts.In this thesis, the composition of automated fingerprint identification system was introduced based on the technology processes clues. The studies focused on the image pre-processing and feature matching two processes, image pre-processing section was focused on the pattern of the orientation estimation and image segmentation algorithm, related algorithms were improved. The orientation estimates discussed the direction number of neighborhood direction template. Block pattern calculated using the templates and direction histogram combination. Fingerprint image segmentation combined the variance method and direction method. The segmented image contours were smoothed by using expansion and corrosion of the morphology Feature extracting section extracted the overall characteristics of the diverging points and the local characteristics of bifurcation points and endpoints information. Then used a point set form to represent the extracted feature points. Feature matching part were based on large number of fingerprint template library cases. Used the fingerprint global features to classify the fingerprint image in the first, and then used the singularity information for rough matching, finally used the local feature points for details matching. Local feature points matching based on the point pattern matching method. Proposed a center reference point method for the reference points selection in the pattern matching, then using the reference points converted feature points into the polar coordinate representation. The simulation test of this algorithm using FVC2000 fingerprint database and the experiments show that the algorithms we mentioned above can be effectively used for fingerprint identification systems.