An inverse scattering approach using hybrid PSO-RBF network for microwave imaging purposes

This paper presents a new microwave imaging approach for the reconstruction and the dielectric characterization of physically inaccessible cylindrical objects. Due to its superior performance, radial basis functions network (RBF) is adopted to solve the inverse scattering problem. However, the problem of selecting the appropriate number of basis functions and defining the optimal parameters remains a critical issue for RBF network. Hence, it is useful to select an optimization process of hidden layer of RBF network. For better convergence, error rates and object reconstruction results, Particle Swarm Optimization (PSO), a new promising algorithm, is proposed to train RBF network. The incorporation of PSO in RBF Network is accomplished by optimizing the hidden unit number and their centers. In this work, a formulation of the approach is described and images of reconstructed circular cylinder are reported.