Fingerprint Alignment Based On Help Data And An Algorithm Of Extracting Triangulations For Fingerprint Indexing

With the development of society and the growth of economy, the personal identification systems are widely used to guarantee the security and accuray. The traditional methods of using token or password are not reliable for they are easy to be lost, forgotten or illegally used. The technology of biometric identification shows us a more convenient and reliable way to identify an individual by recognizing one's characteristics of biometrics and behaviors. The technology of biometric identification is based on one's biometric characteristics, using technology of information processing,and it is the combination of biometric technology and information technology. In all, fingerprint recognition becomes a most efficient way to identify an individual.In this thesis, we focus on two problems in fingerprint identification, including fingerprint alignment using help data and extracting triangulations for large-scale fingerprint database indexing. Some details are as follows:(1) At present, the algorithm of fingerprint alignment based on minutia point is mature. But in some occasions (like in biometric cryptosystem), the minutia attributes are not available, and only global features in the fingerprint pattern could be used. An approach of alignment based on high curvature points is proposed here. First, through extracting the orientation field for the given fingerprint, orientation field flow curves (OFFC) based on the estimated orientation field are obtained to, extract high curvature points to generate help data. At last a trimmed iterative closest point (ICP) algorithm is used to determine the alignment between the template and the query based on these helper data.(2) With the expansion of fingerprint database, an efficient indexing approach is required. Normally, local structures of fingerprint minutia are used to form the index in the fingerprint coarsely matching process, we choose triangle as the local structure in this thesis, and extract triangulations by establishing a net of Delaunay triangle s for large-scale fingerprint database indexing. Experiments show that, this algorithm of extracting triangulations has a remarkable effect on reducing the number of triangulations.(3)A matching subsystem of an automated fingerprint identification system is bulit. It includes a comparetion and result inquiry...