A COMPARATIVE PERFORMANCE ANALYSIS OF VARIOUS CLASSIFIERS FOR FINGERPRINT RECOGNITION
In this study, recognition of fingerprint images has been performed by recent classifiers as well as some important and common classifiers available in the literature. The classification methods used in the study are support vector machines, k-nearest neighbors, Naive-Bayes, decision tree learning, and deep neural networks. Training/testing data set has been obtained basically by using four different versions of fingerprint images of 165 different fingers. Additional seven rotated versions of each different fingerprint images are also used to extend the data set. Feature vector of each fingerprint image (a fingercode) has been produced by using directional Gabor filters and averaging specific regions (sectors) of their output images. After creating fingercode data set, all classifiers has been trained to recognize fingerprint images. Detailed simulation results show that deep neural networks can be effectively used among all classifiers for recognition of fingerprint images.
Tam Metin:PDF (English)
- Şu halde refbacks yoktur.