Microwave imaging of dielectric cylinders from experimental scattering data based on the genetic algorithms, neural networks and a hybrid micro genetic algorithm with conjugate gradient

The application of three techniques for the reconstruction of the permittivity profile of cylindrical objects from scattered field measurements is studied in the present paper. These approaches are applied to two-dimensional configurations. After an integral formulation, a discretization using the method of moments (MoM) is applied. Considering that the microwave imaging is recast as a nonlinear optimization problem, a cost functional is defined by the norm of a difference between the measured scattered electric field and that calculated for an estimated relative permittivity distribution. Thus, the permittivity profile can be obtained by minimizing the cost functional. In order to solve this inverse scattering problem, three techniques are employed. The first is based on a basic real coded genetic algorithms (GAs). The second is a hybrid technique (mGA-CG) which is based on a conjunction of a micro genetic algorithm (mGA) approach with the conjugate gradient based method (CG). The third is an application of an artificial neural network (ANN) having multilayered perceptrons architecture (MLPs). Three algorithms: conjugate gradient with Polak–Ribiere updates (CGP), Levenberg–Marquardt (LM) and gradient descent (GD) are used to train the ANN. Computer simulations of these methods are performed for reconstruction of circular cylinders against laboratory-controlled microwave data.