GPR for Tree Roots Reconstruction under Heterogeneous Soil Conditions
Résumé
In the present chapter, a processing framework for tree-root reconstruction under heterogeneous conditions using GPR is proposed and investigated. The core of this GPR data processing is based on the matched filter as a wave migration technique. Heterogeneous soil conditions then make GPR processing challenging. The framework has been tested considering a heterogeneous medium based on an FDTD electromagnetic simulation scenario. The SVD-based clutter removal approach has been chosen since it clearly outperforms other techniques and it offers flexibility in clutter reduction. There exist several solutions to reduce the effect of background noise, mean/median subtraction, high-pass filtering, shifted and scaled background removal and principal component analysis (PCA) approaches. The relative permittivity of the heterogeneous soil itself is inaccessible. But GPR is a delay-based detection technique where root depth is to be decided for a known relative permittivity of the media of propagation. So, during in situ GPR surveys, operators perform additional measurements in order to assess the soil's relative permittivity by empirical methods like borehole GPR, average envelope amplitude, frequency shift methods, or full-wave inversion techniques. Otherwise, methods using dedicated sensors are based on a soil moisture analysis. Indeed, using models such as Topp’s one, the measured soil moisture can be translated into a relative dielectric constant. However, alternatives do exist. Hyperbola fitting is a processing method that can be employed to extract information about the propagation medium. Yielding the relative permittivity of the heterogeneous soil based on GPR hyperbola parameters, the so-called randomized hough transform (RHT) is proposed. This method is efficient and semi-automatic. Tested with homogeneous soils (the soil permittivity being known), a good correlation between actual and estimated values is reached, which does substantiate its use. The GPR processing steps are illustrated by electromagnetic (EM) simulations in complex scenarios. A realistic dichotomous root prototype is tested using the FDTD open-source simulations environment gprMax software. Heterogeneous soils can be investigated by examining the effect of increasing soil moisture with parametric compositions (sandy and clay soils). To provide experimental insight, controlled laboratory experiments using an in-house bi-static GPR system are then run for a simplified 3D root structure reconstruction from GPR data.