Accelerated fingerprint identification method

The article outlines a method for accelerated identification of fingerprint images based on templates as image models. They are formed as a result of automatic processing of images. The method is based on the properties of the nearest neighborhoods of minutiae in the form of endings and bifurcations and consists of two stages. At the first stage, each minutia of the query template is compared with each minutia of the reference template from the database and the similarity of such pairs of minutiae are estimated. To speed up computational operations, classes are introduced that allow you quickly accumulate the similarity of minutiae from these two templates in a histogram. Histograms are built for all reference templates from the database and one query template. At the second stage, based on histogram estimates, the most similar templates are selected, the number of which is much less than the size of the database. These templates are compared additionally taking into account the consolidation of minutiae and the compactness of the location of the corresponding pairs of minutiae. Significant acceleration of the identification algorithm is achieved by discarding dissimilar pairs of minutiae at the first stage and pairs of patterns with poor histogram estimates at the second stage. The results of experiments are presented, which are published on the Internet.