An efficient approach for the solution of inverse scattering problems via neural network method

In this paper, an efficient approach to the solution of inverse scattering problem related to circular hole buried in a wall is presented. The approach is based on neural network method in which the reconstruction domain is first restricted to a narrower one by Electromagnetic Back Propagation Method (EBPM), and then classical neural network method is applied. Comparisons between the classical and proposed neural network solution of the problem show that the present method can detect the circular crack more accurately than classical solution for the same number of education set element. Also proposed solution provides faster education than classical solution by means of restriction. The approach may also be used to accelerate the neural network based methods for more complex inverse scattering problems.