Research On The Intelligent Fingerprint Matching Algorithms By Combining Global And Local Features

Automated fingerprint identification system (AFIS) is a technology of fingerprint identification using computer, it is of convenience, high efficiency, security and reliability. It has been applied in many fields such as financial security, data encryption, electronical business and will play a more and more important role in our life.This paper is mainly about fingerprint feature extraction and matching algorithm. The work focuses on the following aspects:We adopt an improved algorithm for minutiae extraction, without trivial ridges restoration on thinned fingerprint images. First, the set of original minutiae is directly extracted from thinned fingerprint images by the way of template detecting method. Then, analyzing various noises in fingerprint images and their properties, generalizing the distributing regulation of pseudo minutiae, combining with the information of local ridge direction to design special algorithms for deleting pseudo minutiae from the original minutiae set.The paper points out the advantages and disadvantages of GA(Genetic Algorithm) and BP(back-propagation) algorithm, researches a hybrid intelligent learning algorithm that combines GA with BP algorithm for fast & exact neural network training. The simulation results show the algorithm has high convergent speed, easily-oriented global optimization and practical value.An intelligent fingerprint matching algorithm combining minutiae with the orientation field is proposed in this paper. In the procedure of local feature matching, the paper adopts an improved local vector matching method for point pattern matching and seeking the reference points. The local vector is formed of ridge bifurcations which have high reliability; This paper uses orientation fields as global features, aligns two fingerprints'orientation images according to the reference points, then regards the consistency of the orientation fields as global matching score. Finally, this paper trains the neural network using GA and BP combination algorithm, the inputs of the network are the matching score of minutiae and orientation fields respectively, the output is the combined matching score(0 or 1). Fingerprint recognition uses a proper threshold to make a decision.Experiment results show that the intelligent fingerprint matching algorithm which combines global and local features can meet the needs of common password entry.