Research On The Key Algorithms Of Automatic Fingerprint Identification System

The fingerprint identification technique was deep learned and applied extensively in thecreature examination. People have make use of fingerprint in personal identification for a fewcenturies. With the development of science and technology, the fingerprint identificationtechnique also gets the homologous development, the Automatic Fingerprint IdentificationSystem (AFIS) is a system, which is to carry out the automation of fingerprint identificationwith computer.Each key algorithm involved in AFIS was studied in this paper, several tasks were done:1) Rough set theory was imported to partition of fingerprint image, the image wasdivided with block gray and block grads. Then using the positive area, negative area andboundaries to certain its adscription. The foreground of fingerprint image was able to bedivided correctly. The speed was also very quick (0.06s or so).2) Ridge endings and bifurcations were extracted in the thinning image according to theminutiae's property. An algorithm was designed to eliminate pseudo-characteristics in everystatus, which was proved effective. An algorithm was put forward to change block direction tominutiae direction. Finally adding the surrounding Gabor values and frequency values tocreate fingerprint character database.3) A fingerprints matching algorithm based on genetic algorithms was put forward. Thecoding scheme and copy, crossover, mutation strategies were designed, which could realizefingerprints matching quickly (0.16s or so). At the same time, "combination bombing" wassolved. This algorithm doesn't depend on fingerprint center and is insensitive to displacementand rotation. The error rate is about 11.2%. When false acceptance rate is 0%, the falserejection rate is about 18%. These algorithms are preferred to the AFIS, which doesn't acceptthe false fingerprint.