Research on fingerprint classification based on twin support vector machine

Abstract

Fingerprint classification is one of the core steps of fingerprint recognition and directly relates to the accuracy of recognition. For this reason, a fingerprint classification method based on Twin Support Vector Machine (TWSVM) is studied. First, the Gabor filter is used to extract texture features from fingerprint images. Second, a multi-class model based on TWSVM is constructed by using the 'one-versus-all' strategy and the binary tree method, respectively. The quantum particle swarm optimisation algorithm is used to optimise the parameters in the model. Then the fingerprints are divided into five categories using the optimised model. Finally, the classification model is evaluated using fingerprint images from the NIST-4 database. The experimental results show that applying the TWSVM to fingerprint classification can get good classification results.