|
Acadêmico(a): Rafael Araújo Gumz |
Título: Protótipo de um Sistema de Identificação de Minúcias em Impressões Digitais Utilizando Redes Neurais Artificiais Feedforward Multicamada |
|
Abstract: |
The actual monograph introduces a prototype where an artificial neural network multilayer feedforward was trained with samples of pieces from fingerprint raster images that contain three types of minutia. Those types are divided in ridge bifurcation, ridge ending and false minutia. Then, from this neural network was taken input data of minutia to classify those inputs and show in the fingerprint image where the identified minutia was found. Yet, in this monograph the construction of a part of an automatic fingerprint identification system is related, a totally functional prototype for minutia classification is presented, the results of backpropagation training are shown, and finally the difficulties involving the use of this sort of neural network applied to the minutia classification problem are discussed. |
|
|