ARTIFICIAL NEURAL NETWORKS FOR SOLVING OF OPTICS INVERSE AND DIRECT PROBLEMS

摘要

Possibilities of artificial neural networks (ANN) to solve problems of determination of local characteristics (an inner structure of an object) and integral characteristic (estimation of an object as a whole) in a case of incomplete information about object image (or an object signal) are presented. It is very important when an object cannot be visualized (registered) as a whole as well as in a number of other cases when ordinary methods of solution of inverse problems and direct tasks can not be used. Possibilities of using of the only one (one-point measurement) or two (two-point measurement) values of a signal distribution function in a plane of the registration in order to determine a full distribution of local characteristics in an object with a cylindrical symmetry as well as its integral characteristics are shown. In a case of homogeneous object, ANN allow to define a form, size and orientation of an object. In prospects, ANN can be used for creation of automated systems for diagnostics, testing and control in various fields of scientific and applied research as well as in industry.