Research On The Automatic Fingerprint Identification System

Fingerprint identification technology is one of the most widely used identification methods.Fingerprint identification technology solves the problem that traditional code easy to lose andforget that may cause economic lose.This paper researches the fingerprint image segmentation enhancement, binarization,thinning, feature extraction and matching. And the Research is mainly focused on the fingerprintimage binarization, fingerprint image feature extraction and fingerprint imagematching. Analyze some algorithms and improve them, so it can improve the speed andaccuracy of fingerprint identification system.In order to make the binarization fast, the paper put forward a based on binary particleswarm optimization (pso) algorithm of information entropy Binary Conversion method. Theexperiment shows that it makes the binary-conversion fast and can receive great result.In order to choose a more suitable thinning method for the fingerprint system, this papercompares the improved OPTA thinning method and refining the look-up table method. Theexperiment result shows that the refining look-up table method is better than the OPTA thinningmethod because the improved OPTA method may make burrs into the fingerprint and thethinning effect of look-up table method is better than OPTAmethod.For the fingerprint feature extraction is too slow and feature many points are inaccurate, thispaper applies support vector machine (SVM) fingerprint feature points extraction, in order toachieve the purpose of improving fingerprint extraction speed and quality. when remove the falsefeature points, the method not only like traditional methods considering the distance relationshipbetween feature points, but also consider the fingerprint direction field of relationship betweenfeature points and makes to it more accurately.Traditional matching methods often need to compute calibration function to alignment. It istoo difficult to implement, so this paper designs a simpler method of matching. This method isbased on the relatively center distance and direction of the feature points and the type of featurepoints, experiments prove that this method is effective to overcome the effects of the fingerprinttranslation and rotation.