Decision on the identity of the persons on two-dimensional images using a neural network approach

The paper presents an approach used in solving the problem of identification of a person's face in his two-dimensional image. To train the neural network requires a large number of images of identifiable persons, and they are usually there a little bit. In [1] it is proposed together with an actual images to use simulated training material. But it is necessary to have a reliable way of automatically making a decision about the identity of the persons on actual and simulated images. It is proposed to make a decision about similarity or dissimilarity of persons put to the input of the neural network is not the images themselves, and special fields, reflecting the geometric differences compared images. When training a neural network uses the field of two classes: one estimated from pairs of images of the same person, and the other pair of images of different persons.