Multi-modal Identification System Based On Face And Fingerprint Studies

Modern society, with the rapid development of information technology, the exponential increase in the amount of data, more and more new technology come to people's daily lives, people can feel more and more warm and convenient which digital life has brought us. Early on, in order to identify a person's identity, the traditional approach is to use a password or a card for authentication, this traditional technique solve the fundamental question of identity in certain extent. Because the traditional identification techniques have inherent defect, people are turning to research what is based on biometric identification technology, compared with traditional technologies, this new technology can effectively compensate for its flaws and shortcomings. biometrics also have good uniqueness, stability and extensive. In recent years, more and more biological characteristics are used in the biometric identification technology, for example, fingerprint, face, iris, etc., as well as hand-type, palm, signature, voiceprint and more. So many biological characteristics can be used, therefore, a lot of researchers have turned their attention to multi-modal identification technology research in this field. In this thesis, based on the research of face recognition technology and fingerprint recognition technology, fuse more than two kinds of modal biometric.In the thesis, carried out the following aspects of research work:(1) In the study of face detection, face detection technology based on adaboost was modified to achieve a algorithm that by locate the position of human eye and face, accurately detect the region and the attitude of human face, the algorithm improve the normalization effect after face detection.(2) In this thesis, face recognition technology is studied deeply, implemented four face recognition algorithms based on PCA features, LDA features, GABOR features, LBP features separately, the algorithm parameters for the experiment were adjusted repeately to achieve good results.(3) For fingerprint recognition technology, fingerprint preprocessing methods are discussed in this thesis, including the direction filtering, binary image segmentation, fast thinning. On the binary thinned image, discussed that how to extract and save detaile features such as bifurcation points, and how to remove the false feature points which have been extracted, and the feature extraction and matching method were systematically studied, describes the local characteristics of the fingerprint minutiae feature points vector and the data structure, the first matching processing, the adjustment of the coordinates in the second match, and global matching and matching conditions judgments.(4) The details of the basic theory of information fusion, and fusion method of three modes feature that the LDA features, GABOR features of face, and the fingerprint feature are studied, achieve GCCA-based feature-level fusion by LDA features and GABOR features, and then use the fusion feature of facial features and fingerprint feature to fuse in match-level, by adaptive weighted and D-S evidence theory that two match-level fusion methods respectively, get a better fusion result.(5) Introduce the design of a multi-modal identification system briefly, Describes the system structure and parameters related to the design, including databases, file structure design, and the specific function of the three modules structure. An extended class design method also was introduced to improve the system robustness and stability of the platform.