Fast regularized linear sampling for inverse scattering problems. (Q2889369)

The linear sampling method has recently been the object of a growing interest in inverse scattering problems. In the paper, the authors present several modifications of a published computational procedure designed to improve its efficiency and remove its hand-tuning part.NEWLINENEWLINEThe authors define the direct as well as inverse problem to be solved by the linear sampling method. From field measurements, an approximate far-field pattern is known for \(N\) incident plane waves of different directions. At each space point, we finally obtain a system of \(N\) linear algebraic equations for the approximate solution, where \(N\) may be very large. The inverse problem is ill-posed and the Tikhonov-Morozov regularization is applied.NEWLINENEWLINEAll the systems have the same dense matrix whose suitably truncated singular value decomposition is computed iteratively and used to solve the individual systems. Finally, the fact that the gradient of this solution is large on the boundary of the object is used to reconstruct the object. Five numerical examples with \(N=2252\) and \(50\times 50\times 50\) points show the applicability of the method proposed to solving inverse acoustic scattering problems. The time necessary for the computation is shown, too.