A linear optimization method to solve 2D Inverse Scattering Problem with masked domain

Solutions of Inverse Scattering Problems (ISP) are exploited for 2D imaging process, in order to find shape, location and constitutive parameters of targets (e.g. dielectrics) embedded in a electromagnetic scenario. From the mathematical point of view, the ISP is implemented as the minimization of a proper cost function, made up of two terms: they are known as state and data equations, usually both non linear. In the presented work, a linear optimization formulation of the ISP is made, exploiting evolutionary type methods (GA, Stud-GA, PSO and BBO) with the aid of a Contrast Source (CS) formulation. Combining these strategies, the ISP is computed solving the dual Forward Scattering Problem (FSP) for each individual of the starting population and the non linearity issue is demonstrated to be bypassed. A comparison between the four algorithms is performed in terms of speed of convergence when applied to a reference cylindrical geometry, where unknowns has to be searched only in a fixed region, masking the overall domain.