METHOD FOR AUTOMATICALLY RECOGNIZING FINGERPRINTS

• US 20090185724A1
• Filed: 10/31/2006
• Published: 07/23/2009
• Est. Priority Date: 11/08/2005
• Status: Active Grant

• Current Assignee: In-Idt
• Original Assignee: Thales SA
• Inventors: Allouche, Cyril, Marti, Sandra
• Granted Patent: US 8,194,943 B2
• Time in Patent Office: Days
• US Class Current: 382/124
• CPC Class Codes: G06V 40/1347 Preprocessing; Feature extraction

Abstract

The present invention relates to a method of automatically recognizing fingerprints, consisting in establishing a database by digitizing images of fingerprints of individuals, by detecting the corresponding minutiae, by selecting the most discriminating minutiae, by storing the characteristic parameters of these minutiae, then, in the step for recognizing prints of a given individual, in digitizing the fingerprints of this individual, in detecting the minutiae of these fingerprints, in storing their characteristic parameters, in comparing these parameters with those stored in the database, and it is characterized in that, on establishing the database and on taking prints of said given individual, for each print in the database and of the individual concerned, at least the spectra of the selected minutiae are stored as characteristic parameters of the minutiae, and in that, after comparison of the characteristic parameters of the prints of the individual with the corresponding parameters of the prints in the database, a score is deduced for each of the duly performed comparisons, and the decision is made

Claims

1. A method of automatically recognizing fingerprints, comprising the steps of:

establishing a database by digitizing images of fingerprints of individuals, by detecting the corresponding minutiae, by selecting the most discriminating minutiae, by storing the characteristic parameters of these minutiae, then, in a step for recognizing prints of a given individual, in digitizing the fingerprints of this individual, in detecting the minutiae of these fingerprints, in storing their characteristic parameters, in comparing these parameters with those stored in the database, wherein, on establishing the database and on taking prints of said given individual, for each print in the database and of the individual concerned, at least the spectra of the selected minutiae are stored as parameters characteristic of the minutiae, and in that, after comparison of the parameters characteristic of the prints of the individual with the corresponding parameters of the prints in the database, a score is deduced for each of the duly performed comparisons, and a decision is made.

2. The method as claimed in claim 1, wherein a decision is made by comparing the final score, obtained from all the comparison results, with a threshold determined by learning.

3. The method as claimed in claim 1, wherein the calculation of the spectrum of each minute detail is optimized by extracting a box encompassing this minute detail to obtain a thumbnail image of the content of this box, by calculating the spectrum of this thumbnail image, by realigning this spectrum by rotation according to the orientation of the minute detail, by reducing the dimension of this spectrum after rotation and by storing this reduced spectrum.

4. The method as claimed in claim 1, wherein the spectra of the minutiae are obtained by Fourier transform.

5. The method as claimed in claim 3, wherein the reduction of the dimension of the spectrum is done by selecting a useful spectral band.

6. The method as claimed in claim 3, wherein the reduction of the dimension of the spectrum is done by main component analysis.

7. The method as claimed in claim 3, wherein the reduction of the dimension of the spectrum is done by undersampling.

8. The method as claimed in claim 1, wherein, in view to calculate a fingerprint comparison score, the following calculations are performed:

the probability of matching each pair of minutiae by calculating the correlation C of their two spectra, and then the probability P of matching this pair, which depends on the correlation C and the valencies of the two minutiae.

9. The method as claimed in claim 1, wherein the comparison is done only using the spectra of the minutiae.

10. The method as claimed in claim 1, wherein the comparison is of the “

Hough matching”

type and is performed using the spectra of the minutiae, and the coordinates, orientation and valency of the minutiae.

11. The method as claimed in claim 1, wherein the comparison comprises a first step of the “

Hough matching”

type performed using the coordinates, orientation and valency of the minutiae, this step including a calculation of minute detail comparison and matching scores, and a second step for comparing spectra of the minute detail matches obtained in the first step.