Fingerprint Recognition Algorithm

While a significant progress has been made in the research and development of automatic fingerprint recognition, the application of this technology does not prevail nowadays. The reason is that the accuracy and speed of the recognition is far from satisfactory to many practical circumstances. To improve the performance of automatic fingerprint recognition is necessary both in theory aspect and application aspect. For this reason, this thesis has discussed the problems of fingerprint segmentation, fingerprint enhancement, fingerprint matching, and fingerprint classification. The major contributions of this thesis are listed as follows:1. Through analyzing the existed algorithms for the computation of fingerprint orientation fields, a method to compute pixel orientation field and block orientation field of a fingerprint is proposed. It can accurately compute the orientation fields based on the structure tensor method.2. Two novel algorithms for fingerprint segmentation are introduced: a method based on gradient projection and morphology, and a method based on PCNN and morphology. Both methods have adopted the morphological operation, so the contour of the extracted fingerprint foreground is much smoother. In addition, both methods do not need to calculate the texture features, so the computation amount is much less.3. Based on 2-D Gabor filtering enhancement, a cascade fingerprint enhancement method based on 1-D filters has proposed. The cascade method means enhancing the fingerprint image from coarse to fine: enhancing the fingerprint image from using the block orientation field to using the pixel orientation field. Moreover, using two 1-D filters to replace the 2-D Gabor filter will speed up the calculation of fingerprint enhancement.4. Based on bar-model developed by two minutia points, a fingerprint matching method based on the minutiae's geometrical and textural (angular histogram around the minutiae) features is demonstrated. After adding the angular histogram around the minutiae as the textural feature, the reliability of minutiae's matching is improved, and the matching efficiency is greatly improved as well.5. A novel fingerprint classification method is illustrated, which has adopted the singular points and angular histogram of the orientation field as the features for classification. Meanwhile, it has adopted the BP-network and the SVM as the pattern classifier. Using singular points can efficiently identify the arch pattern, and using the angular histogram can efficiently identify other patterns.