Reconstruction of the Characteristics of a Light-Scattering Layer from Its Reflectance and Transmittance: The Neural Network Method

Abstract

The problem of reconstructing the probability of photon survival, the optical layer thickness, and the parameter of asymmetry of the scattering indicatrix of an elementary volume from the dependence of the reflectance and the transmittance on the angle of incidence of radiation onto a layer of a medium with Fresnel reflection from the boundaries is considered. To solve this problem, a method based on the use of correlation neural networks is proposed. For training and control of the operation of neural networks, the reflectance and the transmittance of the layer calculated within the framework of the radiation transfer theory were used. For the simulation of the scattering characteristics of an elementary volume, the Mie theory was used. Estimates of errors in the reconstruction of the characteristics of the medium were performed.