Incomplete Fingerprint Matching Method

In recent decades, biometric authentication technologies have drawn much attention due to their many advantages such as automation, safety, reliability and convenience. They are playing more and more important roles in modem human life. All of them, fingerprint recognition is the most popular one. Though many successful algorithms for fingerprint recognition have been proposed, there are some challenging problems remained which can strongly influence the performance of fingerprint recognition systems, such as feature extraction and matching of partial fingerprints.The research work in this thesis is focused on feature extraction, graph representation and matching of partial fingerprint images. The main contributions are summarized as follows.(1) The concept of Representative Ridge Points (RRP) is introduced in the stage of feature extraction. After RRPs are extracted from ridge skeletons, the local structures Improved K-plet (ImKplet) are created by both minutiae and RRPs as local representation of fingerprints.(2) We present a global ImKplet-based graph representation of fingerprints. In this model, the constructed graphs can cover more regions of fingerprints which benefit the following matching procedure and improve its performance considerably, especially for partial fingerprints.(3) We present a novel method Improved CBFS (ImCBFS) for fingerprint matching. Owing to the advantages of ImKplet graphs, the method can match the input and template fingerprints in a broader region consistently which leads to more reliable matching results.(4) Lots of experiments are conducted on the public databases of FVC 2002, and the experimental results are inspiring to show that the proposed algorithm in this thesis has superior performance than the original K-plet/CBFS method.