Microwave Imaging of Partially Known Targets Through a Mild Data-Driven Approach

Microwave imaging is a powerful tool in many applications where non-destructive and non-invasive inspections are requiredsince the adopted frequency range allows the use of relatively cheap and safe apparatuses. However, the image reconstruction process, especially when a quantitative reconstruction of the dielectric properties of the target is required, involves the solution of an inverse problem, which is notably non-linear and ill-posed. In this framework, different techniques have been proposed in the last years, ranging from stochastic to deterministic solving schemes [1]. In several cases, the imaging problem is however related to the identification of defects and inclusions inside (partially) known host structures. In such conditions, the inclusion of some a-priori information into the inversion procedure may allow a significant enhancement in the retrieved solution.