| Due to its uniqueness, reliability and permanence, fingerprint has become the most important symbol of individual identification, which also was acknowledged as"the chief physical evidence". Although the research and development of AFIS (Automatic Fingerprint Identification System) has obtained significant progress, some difficulties still exist. For example, the accuracy of one-to-one identification between fingerprints which were collected under different environments is not ideal; and the performance of verification in large-scale fingerprints databases is far from satisfactory in practical circumstances. Address these issues, the essay mainly studied fingerprint image quality estimation algorithm and fingerprint matching algorithm, using the knowledge of digital image processing, intelligent algorithms, etc. The major contributions are listed as follow:(1) The critical techniques in AFIS, including quality estimation, preprocessing, feature extracting and matching algorithms, have been implemented. These works lays the foundation for the experiments of the followed researches.(2) Analyzing the shortage of existed fingerprint image quality estimation algorithms, and try to use neural network to estimate the quality of fingerprint image. This algorithm takes local features of fingerprints as the inputs into neural network and quality score as output. Neural network is used to learn the estimations of image quality of human being. Relative to other algorithms which are based on local features, this algorithm can estimate this image quality more accurately.(3) Shape context algorithm is applied in fingerprint matching process to improve the performance. Shape contexts can capture the relative distribution of minutiae in the plane relative to each minutiae in minutiae set. The essay improves the original shape context algorithm, which adding features of fingerprints, so that the algorithm can be suitable for fingerprint matching. |
PDF Full Text Request