| Accurate, secure and practical personal identification methods are highly required with the development of the social and economy. The traditional personal identification methods based on token or password have the hidden troubles. The token or password is prone to being lost, forgotten, copied and stolen. The traditional personal identification methods cannot satisfy the security requirement of our highly interconnected information society. Biometrics based on the physiological or behavior traits identification provides a convenient and reliable scheme. Among the numerous biometrics, more attentions have been paid to the fingerprint identification technology due to its convenience, high accuracy and low cost. The technique of fingerprint identification has become one of the widest used biometric identification techniques. It has been widely used in electronic commerce, criminal identification, information safety etc.Fingerprints are scraggly lines on the tip of fingers, which are invariable in one's whole life after they formed in the early times of a person's life. The automatic fingerprint identification system (AFIS) is a synthetical system that integrates photoelectric technique, image process, computer and network, database technique and pattern identification technique. It commonly includes fingerprint image collection, fingerprint image process, minutiae extraction, minutiae matching etc. Aimed at the running process of the AFIS, some key algorithms used in AFIS are studied in this paper based on summarizing existed research results. The paper presents effective algorithms in the aspects of the synthesis evaluation and enhancement of fingerprint image, fingerprint classification, fingerprint identification, etc. The AFIS designed in this paper has been used in the on-line identification system of retirement pension payoff successfully.The main points are as follows:(1) The basic concept of the biometric identification system and thefamiliar biometric identification techniques are summarized. The structure, principle and the research status quo of AFIS are introduced. The key problems in AFIS are analyzed.(2) The preprocessing of fingerprint images is studied, which includes the synthesis evaluation of image, fingerprint enhancement, binarization, thinning and repair. Based on the study of existed image evaluation methods, a practical synthesis evaluation method of fingerprint image is proposed. Different evaluation strategies are applied in fingerprint registration and identification process. Rough evaluation and fine evaluation are used in registration process in order to ensure the reliability of fingerprint template, and rough evaluation is adopted in identification process. In the process of fingerprint image enhancement, the fingerprint image enhancement algorithm based on adaptive filter and fingerprint image enhancement algorithm based on the curvelet transform are presented in order to improve the definition of fingerprint ridges and eliminate the noises effectively. Compare of different algorithms is carried out. An expert-experience-based fingerprint image repair algorithm is presented in order to eliminate the burrs, bridges and complex connections in the skeleton image after thinning. Burrs, bridges and complex connections on the ridge are processed separately by improved angle calculation method and the ridge length threshold. The slope method is used in the repair of break ridges.(3) Minutiae extraction and classification algorithms of fingerprints are researched. The existed minutiae extraction and classification algorithms are reviewed. A tow-stage fingerprint classification algorithm based on the Support Vector Machine is presented. Using the superiority of SVM to solve tow-class classification problems, we designed a tow-stage fingerprint classifier, presented and realized a novel method about fingerprint classification.(4) The commonly fingerprint identification algorithms based on theminutiae matching are studied. In order to improve the robustness of fingerprint matching algorithm, a contour-based fingerprint identification algorithm is presented, in which contour sets are used as the matching parameters, and so the affect of noises is small.(5) The application of AFIS in the background of the identification system of retirement pension payoff is studied. The background, modeling programming and implementation of the system are expounded. The UML is used as the visual modeling language. The Rational Rose is used as modeling tool. The object-oriented analysis and design ideas are introduced in the on-line fingerprint-verification-based pension payoff system. Rational Unified Process (RUP) is used in the design process. Based on the analysis of the bank system mechanism, a communication protocol conversion device between AFIS and bank system is designed. The flow of pension payment is similar to the general payment in the bank. The retiree can receive the pension payment after pressing fingerprint when he receives the hint to input the password. The flow does not affect other operation of the bank. Thus, the on-line fingerprint identification of pension payment is implemented and the identity verification can be carried in real time. The system has been successfully operated in a region of Shandong province, which solved the imitating drawing problem of the retirement pension to a great extent. The system is of great social benefit, and of great spread value.(6) The work of this thesis is summarized, and the directions of future work are given.According to literature search and science and technology query, the following points have not been reported identically which are the main innovation points of this paper.(1) A practical synthesis evaluation method of fingerprint image is proposed. The parameters of rough evaluation and fine evaluation are introduced. Rough evaluation and fine evaluation are used in registrationprocess. So the reliability of fingerprint template is guaranteed. Only rough evaluation is adopted in identification process to accelerate the identification.(2) The fingerprint image enhancement algorithm based on adaptive filter is proposed and adopted. It is efficient and effective. It was compared with an algorithm based on the curvelet transform, which has the character of orientation. The edges of ridge and valley are enhanced through the modification of the curvelet transform parameters. The experimental results show that the algorithm based on the curvelet transform is effective but is very complicated, and its efficiency is lower than the algorithm based on adaptive filter.(3) An expert-experience-based fingerprint image repair algorithm is used to repair the fingerprint skeleton image. The calculation method of angles is improved. Combined with the ridge length threshold, burrs, bridges and complex connections on the ridge are processed separately. The slope method is used in the repair of break ridges.(4) A tow-stage fingerprint classification algorithm based on the Support Vector Machine is realized. Different characters of singular points are used as the distinguishing rule. Using the superiority of SVM to solve tow-class classification problems, we designed a tow-stage fingerprint classifier, presented and realized a novel method about fingerprint classification.(5) A contour-based fingerprint identification algorithm is presented, in which contour sets are used as the matching parameters. So the affect of noises is small and the robustness is high.(6) The fingerprint identification technique is successfully applied in the on-line fingerprint identification system of retirement pension payment.
|
PDF Full Text Request