Applicability of MUSIC-type imaging in two-dimensional electromagnetic inverse problems

Abstract

The behavior of the multiple signal classification (MUSIC) algorithm is investigated when used to locate small dielectric cylinders of specific characteristics in noise-free and noisy scenarios using the TE incidence. We have made three observations regarding the performance of MUSIC in the two-dimensional TE scenario, which reveal the significance of the choice of signal subspace while employing MUSIC and the shortcoming of traditional MUSIC when used to detect degenerate cylinders (which might be so due to the material or geometry of the cylinders). The detailed analysis of the sources induced on a cylinder and their linear dependency on each other gives a distinct insight into the use of MUSIC algorithm to locate it. A non-iterative retrieval algorithm is provided that is based on the least squares method, which retrieves the electric and magnetic polarization tensors of the small cylinders. The algorithm proposed here is mathematically a simple and direct representation of the physical principles and is applicable to degenerate cylinders as well. © 2008 IEEE.