Involved people: Gilles Bertrand, Francisco Bezerra, Michel Couprie. Associated students: Julien Bourgeois, Alexandre Duverger, Johann Desemery, Jerome Leclerq, Fabrice Mano-Martins, Huynh Hoot Quan (ESIEE).
The goal of this study is to be able to compare a fingerprint obtained by means of a miniaturized sensor (SAGEM) with models stored in a data base, in order to allow an access control (to a room, a computer account, etc). The comparison is based on the detection and the mapping of characteristic points called minutiae: most consist of "forks" in the black or white parts of the image (fig. 18(a)). The stored models consist of graphs having for nodes the minutiae, enriched by information (distances, angles) allowing the mapping.
The data processing sequence breaks up into four principal stages: i) extraction of a graylevel skeleton (fig. 18(b)); ii) extraction of the points corresponding to the significant forks of the skeleton (fig. 18(c)); iii) construction of the "image" graph (fig. 18(d)); iv) tentative of mapping of the image graph with a model graph.
A prototype produced by ESIEE students implements these functions, and gives encouraging results on the test set which we constituted (a hundred fingerprints coming from ten people).
 |
 |
| (a) | (b) |
 |
 |
| (c) | (d) |