Frequency Diversity in a Linear Inversion Algorithm for GPR Prospecting

摘要

A tomographic approach is presented for processing GPR data under a multifrequency/mul-tistatic/multiview measurement configuration. The approach exploits the Born Approximation so that a linear inversion problem is achieved. This allows to analyze the reconstruction capabilities of the solution algorithm and also discuss about the requirements of the number and the spacing of the frequencies needed in order to achieve a nonredundant set of data. It is stressed their dependence on the extent of the observation and investigation domains and the dielectric permittivity of the host medium. The investigation is first performed in the simplifying assumptions of lossless environment so to gain guidelines. Then numerical reconstruction results with model and exact scattered field data, for more realistic geometries, are presented with the aim of confirming the theoretical expectations.