Radial basis function neural network to shape reconstruction of conducting objects

In this work, we propose a microwave imaging technique for the localization and the shape reconstruction of physically inaccessible cylindrical conducting objects from the knowledge of the scattered electric field. The problem is treated with the Method of Moment (MoM) and the radial basis function neural network (RBF-NN) is applied. This linear RBF-NN has both good localization approximation and linear computation complexity with the number of dimension and number of inputs. The shape of the cylinder is represented by a Fourier series while applying MoM, a matrix equation is obtained whose elements are expressed numerically by using discrete points. Based on the boundary condition and the measured scattered field, a set of nonlinear integral equations is derived and the imaging problem is reformulated in to an optimization problem. The use of RBF-NN to solve the inverse scattering problem requires the determination of its architecture parameters. The scattered field data and Fourier coefficients of the cylinder are used for training the RBF-NN as inputs and outputs respectively. To show the performance of the method we finally present and discuss the results obtained by applying the proposed technique on simulated data via numerical examples.