Soil Parameters Retrieval Using a Neural Network Algorithm Trained by a Two Layers Multi-scale Bi-dimensional SPM Model

The overall objective of this paper is to retrieve soil surfaces parameters namely, roughness and soil moisture related to the dielectric constant by inverting the radar backscattered signal from natural soil surfaces. We characterize the soil surfaces and subsurfaces by a two layer geo-electrical model. The upper layer is described by its dielectrical constant, thickness, a multi-scale bi-dimensional surface roughness model by using the wavelet transform and the Mallat algorithm, and volume scattering parameters. The lower layer is described by its dielectric constant and multi-scale surface roughness. To compute surface, subsurface and volume scattering, we consider a two layers multi-scale bi-dimensional Small perturbations model. In this study, each surface of the two layers surface is considered as a band limited fractal random process corresponding to a superposition of a finite number of one dimensional Gaussian processes each one having a spatial scale. We investigated the dependence of backscattering coefficient on roughness multi-scale parameters and soil moisture parameters for different incident angles by a sensitivity analysis. This sensitivity analyses is the first step of an inversion procedure. To perform the inversion of the small perturbation multi-scale scattering model (MLS SPM) we used a multi-layer neural network (NN) architecture trained by a backpropagation learning rule. The inversion leads to satisfactory results with a relative uncertainty of 8%.