Fingerprint Singular Point Detection And Pattern Classification Based On Block Pattern

Because of the advantages such as stability, accuracy and safety, automatic fingerprint identification technique has always been a hotspot in recent years and become more and more mature. However, the cost of fingerprint recognition is huge since the fingerprint database has numerous amounts of fingerprint samples to be compared with. Currently, the best way to speed up the algorithm and reduce the workload is fingerprint classification, which provides a working method that fingerprint only has to compare with a specific type of fingerprints. Although a significant progress has been made in fingerprint recognition, the accuracy and speed of existing fingerprint classification methods is far from satisfactory in many circumstances because of the limitation of fingerprint image quality and the complexity of the fingerprint classification.In this thesis, a approach is proposed to improve the accuracy and speed of the algorithm for fingerprint classification system based on singular points. Therefore, fingerprint image acquisition, fingerprint image preprocessing algorithm, fingerprint directional field algorithm, fingerprint singular points algorithm and fingerprint classification algorithm are researched and analyzed in detail in this thesis. The experiment result shows that the proposed algorithm can detect the singular points accurately and effectively achieve fingerprint classification. The contributions of the thesis are as follows:i. In the aspect of fingerprint directional image algorithm, traditional point directional image shows bad robustness to noises and traditional block directional image performs poorly in continuity of block edge. In order to overcome these weaknesses, a enhancement point directional image algorithm based on grouping is put forward uses continuously distributed directional image before the block directional image is calculated. Block directional image based on continuously distributed directional image presents accurate fingerprint characteristics and can effectively solve some drawback of traditional block directional image such as bad continuity and great error of block edge. The result of experiment shows that the proposed algorithm exhibits not only more smoothness, well continuity and gradualness, but also the real direction of fingerprint.ii. In the research on fingerprint singular points detection, due to the existing Poincare index’s limit of losing true singular points and including fake singular points, a modified algorithm for Poincare index is proposed which contains two additional conditions. To some extent, this method enhances the anti noise performance of traditional Poincare index and decreases fake singular points. According to the characteristics of the fingerprint direction in singular points area, a concept of singular points detection based on block directional image search is proposed. This algorithm, firstly, finds a general area where the direction changes severely. Then searches in the general area to narrow the scope of singularities and to precisely locate the singular points. This algorithm can accurately find the core point and effectively get rid of the intrusion of fake points. For delta point, a combination between improved Poincare index and block directional image search algorithm is used to obtain a result with good liability and veracity.iii. In the aspect of fingerprint classification, fingerprint classification is accomplished based on the singular points’number and the relative position of core point and delta point. Compared with other methods, the algorithm of classification based on singular points is simple, fast, effective and reliable.