[Identification of neuronal system objects with standards].

Abstract

At some fixed position of a three-dimensional object in relation to the retina plane let us name the set of coordinates of its characteristical points the standard. It is shown that the projections of the objects to the retinamay be identified with their standard (independent of the objects shift, their rotation, change of the scale and of partial shielding of some objects by the other). The structure of the net system is presented which consists of linear, threshold and integrating neurons where such identification may be performed. The system consists of a net with a feedback for calculating optimal transformations, a net for calculating the degree of proximity between the reflection and the standard and a net generator of hypotheses on the correspondence of characteristical points. All the objects whose standards are presented in the system may be identified simultaneously.