Data Fusion of Ground Penetrating Radar and Electromagnetic Induction for Reconstruction of Soil Electrical Properties

We jointly analyzed the ground penetrating radar (GPR) and electromagnetic induction (EMI) synthetic data to reconstruct the electrical properties of multilayered media. The GPR and EMI systems operate in zero-offset, off-ground mode and are designed using vector network analyzer technology. We compared different approaches for GPR and EMI data fusion. As a first approach, we weighted the EMI and GPR data using the inverse of the data variance. The ideal point method was also employed as a second weighting scenario and the third approach is the naive Bayesian method. Synthetic GPR and EMI data was generated for the particular case of a two-layered medium. Analysis of the objective function response surfaces from the two first approaches demonstrated the benefit of combining the two sources of information. However, due to the variations of the GPR and EMI model sensitivities with respect to the medium electrical properties, the formulation of an optimal objective function based on the weighting methods is not straightforward. While the Bayesian method relies on assumptions with respect to the statistical distribution of the parameters, it may constitute a relevant alternative for GPR and EMI data fusion.