Fingerprint Algorithm And Realization

Fingerprints are the ridge and furrow patterns on the tip of the finger and are used for personal identification of people. The pioneering studies in this field examined the details that reside in fingerprints and the global structure of fingerprints. However, fingerprint identification is still a challenging research problem. Comparing the verification performance, both the accuracy and the speed may decrease significantly if a verification algorithm is simply extended to solve an identification problem. For these reasons, this thesis has discussed the problems of fingerprint image-quality estimation, fingerprint classification, fingerprint segmentation, fingerprint enhancement, and fingerprint matching. The major contents of this thesis are listed as follows:1. Two algorithms for fingerprint image-quality estimation are introduced:a method based on direction and minutiae, and a method based on feature fusion. The former used the details of the local characteristics of the fingerprint, and the later used the global features of fingerprint.2. Fingerprint classification and segmentation. First analyzing and testing a novel fingerprint classification, which using genetic programming to extract the fingerprint features, meanwhile adopting the BP neural network and support vector machine as the pattern classifier. However the higher is the complexity of the algorithm, the efficiency is low. Then analyzing and researching a fingerprint segmentation algorithm which is based on Cellular neural network and morphological algorithm. Cellular neural network can take into account the pixel neighborhood correlation in space, this feature is very suitable for fingerprint segmentation and extraction, and the edge of the fingerprint image is relatively smooth by making use of the morphological algorithm. However the algorithm accuracy and efficiency is not high.3. Analysis the second-generation Curvelet transform based enhancement method. The second-generation Curvelet transform can effectively improve the poor quality of the collected fingerprint image texture, and have the strong ability to repair the fracture gap between the ridges. All the algorithms are simulated through MATLAB7.0 simulation platform, the test results showed that classification algorithms can not meet the needs of real-time systems, segmentation showed no superiority of the existing algorithms, only the enhancement is better, but some of the enhancements into the C code running on the system in the original system, although the test results slightly better than the original, more time-consuming.