Inverse scattering by dielectric circular cylindrical scatterers using a neural network approach

Inverse scattering problem has gained a considerable amount of attention in the last decade especially in the field of microwave imaging and object reconstruction. Most of the numerical and analytical techniques used are complicated, ill-posed and require a large computational effort especially when the electrical radius is large as compared to the wavelength. The problem becomes even more complicated when dielectric scatterers are considered. Different numerical methods have been explored in literature to solve the inverse scattering problem. Neural networks are drawing considerable attention as a new paradigm of information processing because of their self-adapting capability in solving problems, more and more applications involving electromagnetic problems are being investigated. A technique based on neural network analysis is presented where the network is trained to model the nonlinear relationship between the dielectric constant and the complex scattering coefficients. The results are verified by applying the technique to a different set of coefficients for a wide range of dielectric constants. The comparisons show that this method is effective and efficient.