Research On The Technologies For Automatic Fingerprint Identification Based On High-capacity Fingerprint Storehouse

Fingerprint identification is a most safe and reliable technology in traditional identification technology, and has been widely used in many fields. However the existing algorithms still show many shortages at speed, accurateness for the identification of large capacity fingerprint storehouse. This paper has taken some deep researches in automatic fingerprint identification system based on large capacity fingerprint storehouse, including fingerprint image quality evaluation, segmentation, enhancement, feature extraction and so on. The main contributions are as following:(1) A fingerprint quality evaluation algorithm is studied and a new method is proposed to evaluate fingerprint quality based on the evaluation of information usability and spectral analysis in this thesis. At first, according to the different fingerprint evaluation factors (effective area, whether singularity, singular point of the migration degrees), evaluate the information usability of fingerprint image hierarchically. If the fingerprint image doesn't meet the requirement, the process of evaluation is ended immediately, and the user is prompted to input fingerprint again. Then the ratios of spectrum energy and the spectrum variance are used to judge the quality and the type of the qualified fingerprints.(2) A two-stage segmentation strategy is proposed in the research of fingerprint segmentation algorithm. At first, the fingerprint image is divided into three categories:blank area, texture area and can not resume area according to the characteristics of the fingerprint block, and then the blank areas and not resume areas are separated through multistage segmentation. The grayscales contrast is treated as the primary segmentation feature to remove the fingerprint blank area and realize the primary division; the block cluster degrees, the number of the consistent of point direction and block direction, and block direction consistency are used as the secondly segmentation feature to remove the can not resume area and realize the secondly division.(3) Fingerprint enhancement algorithm is studied, and an improved enhancement algorithm based on frequency domain filtering is proposed. Firstly, the original fingerprint image is converted to frequency domain through FFT, analyzing the spectrum chart of dry or wet fingerprint, and a passband filter-round filter is designed to remove the correspond low-frequency or high frequency components, enhance the intermediate frequency components. Then the after transformation image is filtered in the frequency domain for eight directions in order to connect broken ridge, remove adhesion ridge, enhance ridge clarity. Finally the enhanced fingerprint image can be got through inverter and ridge combination.(4) Fingerprint feature extraction algorithm is studied, and a hierarchical singularity extraction method is proposed. At first the location of singular point area is found by using fingerprint curvature, then in the singular point area the pseudo singularity identification method combining with Poincare index value and plural filtering is used to locate accurate area. In the aspect of minutiae extraction, a knowledge-based post-processing method is used to remove fingerprint false features.