A Bimodal Biometric Verification System Based On Fingerprint And Hand Geometry

Accurate personal identification is becoming more and more important to the operation of our ever increasingly electronically inter-connected information society. Traditional personal identification methods, such as: key, password, ID card, definitely can not satisfy the increasing security requirements. Due to its uniqueness and stability, biometric technique has become the most promising candidate. In this thesis, we focus on fingerprint and hand geometry verification, these two kinds of biometric techniques, and propose a bimodal biometric verification system based on these two biometrical characteristics. The origin and developing history of biometrics are introduced firstly, then we conclude and compare the advantages and deficiencies of all biometric characteristics, also the application area of each biometric technique is given. The popular topic of current biometrics research—multimodal biometric system is emphasized. Because the fingerprint and hand geometry are utilized in this thesis, a comprehensive literature investigation is done on these two topics. All the key techniques, like: fingerprint image segmentation, fingerprint enhancement, minutia extraction, fingerprint matching, hand salient points extraction and matching, are classified into several classes. According to the requirements of our verification system, the appropriate techniques are picked out. In chapter 3, a bimodal biometric verification system based on fingerprint and hand geometry is specified. After clearing the diagram of the verification system, we divide the whole system into two parts---fingerprint subsystem and hand geometry subsystem, to describe the model algorithm designation. In fingerprint verification subsystem, a segmentation algorithm based Canny edge detector is introduced first, then Gabor filter is applied to enhance the fingerprint image. Following the traditional way of extracting minutia, we use thresholds and morphological thinning operators to get the skeleton of fingerprint image. Here, with the help of tracking fingerprint ridges in thinned fingerprint images, those spurious minutiae connected with ridges are eliminated; later, ridge breaks are repaired according to both foreground and background characteristics. Finally, fingerprint matching is conducted on the Delaunay triangulation, which is constructed on the filtered minutia. For hand geometry verification subsystem, we explain that how to convert an enrolled hand image into a black-and-white image, and how to extract salient points from the hand image. Finally, distance matrix matching is applied to compare two hand vector features. Of course, as a bimodal system, the strategy of fusing the two matching scores of fingerprint and hand geometry to make the final decision is described at the end of this thesis. We compare our experiment results with others'in the existing literature to confirm the advantages of our algorithm, also try to analyze the deficiencies of the system. Finally, we conclude the whole thesis and give out the future plan.