Reconstructing Object Inhomogeneity Parameters from Near-Field Measurements Using Neural Networks

The paper investigates the near field-interaction with a scatterer located in space R². This area is a priority in problems of medical diagnostics and defectoscopy. The process of wave propagation inside various objects is described using the Helmholtz equation. The field is induced by a point source located outside the body. The problem is reduced to the Lippmann–Schwinger integral equation. A two-step algorithm is used to search for inhomogeneities. A neural network approach has been used to filter the values obtained after a two-step algorithm. This problem arises in acoustics, electrodynamics, and flaw detection, as well as in medical diagnostics. When solving the problem numerically, the order of the matrix obtained in the calculation is about 25 thousand elements. Graphic illustrations of the restoration of the function of inhomogeneities within an object are presented. An experiment has been performed demonstrating the features of restoring object parameters using neural networks. The results show the effectiveness of the calculated data autoencoder filtering. A software package for determining the parameters of inhomogeneities within an object has been proposed and pursued.