Research On Fingerprint Classification And Matching Algorithm Using Synergetic Pattern Recognition

With a rapid development of our electronically inter-connected information society, information security requirements have become more and more impending. An accurate automatic personal identification is critical in a wide range of application domains such as Internet security, automated banking, residence safeguard, health care and electronic commerce etc. Biometrics identifies an individual based on her physiological or behavioral characteristics such as face, fingerprint, iris, DNA and voice etc. With its unique advantages, automatic fingerprint identification has become one of the most reliable biometric technologies. In this thesis, two key technologies of fingerprint recognition system are discussed: fingerprint classification and fingerprint matching. An automatic fingerprint classification algorithm classifies a fingerprint into a number of pre-specified categories according to the features extracted from the fingerprint. The classification scheme will build indexing mechanism in the fingerprint database so that we can perform fingerprint recognition more quickly. We first realize the classical fingerprint classification algorithm based on directional field and analyze the experiment data. And then we propose two novel classification algorithms: K-cluster algorithm and synergetic fingerprint classification algorithm. Here the second one is our main point. We propose synergetic classification based on the study of synergetic pattern recognition. The fingerprint samples constructed by cluster process are our stored prototypes; the input fingerprint image is our test pattern. After the dynamic process, test pattern evolves to the most alike prototype. Compared with the other traditional algorithms, our synergetics based classification has the distinguished superiorities in feature extraction layer. The adjoint vectors representing the statistic feature of fingerprint images make global retrieve possible, promote the classification efficiency and deduce the feature extractiondifficulty. In the following part, we introduce the synergetic approach to the fingerprint matching. In a similar mechanism with classification algorithm, we also perform quite well in our experiment.We studied the feature representation of synergetic pattern recognition and pointed out that adjoint vector is feature representation of according prototype. Furthermore, it represents not only image unique feature but also database comparative feature. A hierarchical synergetic fingerprint identification system is defined. In the experiment process, We propose the order parameter choice mechanism, which could predict recognition false by the order parameter value. In the last part, we value our system by the means of hypothesis testing.