Research On Technologies For Automatic Fingerprint Classification And Retrieval Based On Large Fingerprint Database

With the rapid development of network and information technology, the application field of automatic fingerprint identification expands constantly, meanwhile more application requirements are demanded. The existing algorithms aiming at fingerprint identification for large fingerprint database still show many shortages. Based on the research of automatic fingerprint classification and retrieval technology for large fingerprint database, this paper has done some deep researches in the fields of fingerprint quality evaluation, feature extraction, classification and retrieval etc. The main work and contributions of this paper are summarized as follows:1、A new method based on the combination of global and local features is studied to evaluate the quality of fingerprints. The quality of fingerprint images is evaluated hierarchically through different evaluation indices. If the fingerprint image is not qualified, the process of evaluation is ended and the user is reminded to re-enter another one.2、The structural features of the singular area in fingerprint image are analyzed, and a method for localization of singular points which based on the combination of complex filtering and Gaussian-Hermite moments is proposed. Firstly, the singular areas are located by using the distribution of fingerprint curvature, and then the singular points are detected by means of complex filtering and the behavior of Gaussian-Hermite moments. The method can remove pseudo singular points during the process and detect the pair of singular points which is very close to each other effectively, it can also improve the accuracy of localization of singular points.3、 The characteristics of spectrum structure for fingerprint image are analyzed, and a calculation algorithm for the average frequency of ridges based on analysis of spectrum features is proposed. In order to reduce the effect of nonlinear deformation and gray, enhancement and binarization are conducted after intercepting the central region of fingerprint image. Then the average frequency is computed by Fourier transform, it can reduce the computational complexity and improve the reliability of the results. 4、Fingerprint classification algorithms are studied, and a multi-level fingerprint classification algorithm based on grain type, number of ridges between singular points and average frequency of ridge lines in central region is proposed, which is adapt to the searching and matching for large fingerprint database and improve the identification efficiency of the system. The experimental results show it has high search performance and robustness, which can provide a rapid and effective indexing mechanism for large fingerprint database.