Mild Data-Driven Inversion for Microwave Nondestructive Diagnostics of Plants and Trees

The nondestructive diagnosis of pathologies in plants or trees has a great importance in agriculture and green management. In the context of microwave-based diagnostic techniques, a quantitative imaging method is proposed here to obtain two-dimensional reconstructions of the internal dielectric properties of trunks, starting from scattered-field measurements. The method, which aims at solving a nonlinear and ill-posed inverse problem, combines a mild data-driven approach with an inexact-Newton scheme. In particular, the goal of the data-driven methodology, which is adopted here for the first time in this application, is to introduce prior information about the trunk structure, helping the reconstruction process. The proposed algorithm has been tested with numerical simulations involving different working frequencies and dimensions of the considered trunks.