ABSTRACT
The development of imaging methods able to provide information on surveyed areas as reliable as possible, not subject to user’s interpretation, is an important and still open challenge for ground-penetrating radar (GPR) technology [1–6]. Nowadays, microwave tomographic approaches represent a powerful way to achieve such a goal, but they require coping with the solution of a nonlinear and ill-posed inverse scattering problem [7]. Such a nontrivial task entails the need of appropriate regularization methods [8] to restore the well-posedness of the problem and avoid physically meaningless solutions. On the other hand, defeating nonlinearity requires suitable strategies to counteract the occurrence of “false solutions” [4,5], that is, estimated solutions that match the data (within the expected accuracy) but are indeed different from the ground truth. These difficulties are further worsened when, as in GPR surveys, it is not possible to probe the targets from all possible directions, which results in an obvious reduction of the available data and a consequent deterioration of the imaging results. Last, but not least, the existing solution procedures based on global or local optimization strategies are usually computationally demanding and represent a huge drawback when real-time analyses are required [4–6].
