Soil type characterization for moisture estimation using machine learning and UWB-Time of Flight measurements

摘要

Soil moisture measurements are essential in fields such as Agriculture or Civil Engineering. The measurements of moisture via dielectric constant with electromagnetic techniques have become widespread over recent years. In this paper, we propose a new soil moisture estimation method using low cost and small size commercial Ultra-Wide Band (UWB) modules. These modules are capable of very precise measuring Times of Flights (ToF) of signals between two UWB-transceivers. We used this capability for more precise estimation of the soil dielectric constant. Besides Time of Flight, the used UWB-modules enabled measuring of channel impulse responses. We propose as the novelty using an impulse response signal shape for soil type characterization with machine learning, precisely, Support Vector Machine (SVM). So, the basic idea of the proposed method is, first, soil type characterization using machine learning and then, according to soil type, choosing an appropriate moisture-dielectric constant model for more accurate moisture measuring. The main advantages of the proposed method are smaller geometrical dimensions of the measuring equipment, i.e. UWB Radio-Frequency (RF) modules and probes (antennas), due to higher operating frequencies and, as proven with experiments, more accurate soil moisture estimation, especially for organic soils in comparison with equipment based on the conventional Time Domain Reflectometry (TDR) method. (C) 2019 Elsevier Ltd. All rights reserved.