Dorogyi Y. The information technology for biometric human identification by face picture

The thesis is devoted to solving of important scientific and technical problem of developing models, methods and technology for biometric identification for human face image. Oftalmogeometrycal approach to the identification of a person with anthropometric points of eye region, using a smaller number of characteristic features for recognition of human lossless authentication was created. Modification of convolutional neural network structure, which differs from existing with high generalization power of the first layer, fewer variables and high quality recognition was designed. A modification of the learning algorithm for the generalized convolutional neural network based on error back-propagation algorithm with shorten learning time of network was proposed. A modification of the multiclassificator structure through the use of a new set of classifiers and the mechanism of their interaction, which reduced the number of errors of the 1st and 2nd kind through an integrated study of the multiclassification effect was made. The information technology for biometric identification for human face image based on complex mathematical models, methods and approaches, biometric identification from human facial image that can solve together the two classes of problems - search facial images in the database to identify and preliminary control access to objects that are protected was created.