| Due to the uniqueness and invariability of fingerprint, the automated fingerprintidentification is becoming an attractive alternative to the traditional methods ofidentification. However, there is much special noise in fingerprint images, and manyalgorithms are not opened, so an ideal automated fingerprint identification system(AFIS) is still a difficult research subject.In this paper, a new AFIS based on tree classifier is put forward. Firstly, theproposed AFIS classifies the input fingerprints based on the statistical features (singularpoints). Secondly, the fingerprint minutiae are extracted, which consists of computingthe orientation field, filtering, binarization, thinning and other processing procedures.Finally, the fingerprint is identified based on the minutiae matching module. In thepresented AFIS, some new ideas and approaches are brought forward.(1)In the preprocessing, a new method of computing orientation field is proposed。 And based on the obtained orientation field, a new fingerprint image binarization algorithm is presented with dynamic threshold, which has good capability of noise resistance.(2)A modified fingerprint classification method is proposed based on the fingerprint singular points, in which the singular points are extracted with a new Poincare Index method. And the new Poincare Index is obtained directly from the orientation filed. Compared with other classification features, it has advantages of simplicity for computation and suitable for classification.(3)For the minutiae matching, a structure matching algorithm with reference point and a point-pattern matching algorithm based on artificial immune system are developed with good performance.All the above algorithms are implemented with Visual C++. The experiment resultsdemonstrate that the proposed new methods enhance processing effects of fingerprintimages and improve the classification performance. The research achievements haveimportant reference values for the design of an AFIS.
|
PDF Full Text Request