Moisture Estimation for Precision Agriculture through RF Sensing

Abstract

Convenient, non-obtrusive, low-cost, and accurate sensing of fruit moisture content is crucial for the scientific studies of Pomology and Viticulture and their associated agriculture. It can provide early indicators of yield estimation and crop health as well as providing data for food production and precision farming systems. With a focus on grapes, we introduce SING, a scheme that senses grape moisture content by utilizing RF signals but without physical contact with the fruit. In this thesis, we extend the investigation of the theoretical relationship between the dielectric properties and the moisture content of agricultural products to establish a sensing model in the 5 GHz band. To make the work practical, we are first to measure the dielectric properties of grape bunches (not individually as that would be destructive), presenting a unique measurement challenge as internal grapes are hidden. In doing so, we demonstrate that our technique precisely estimates moisture content to a high degree of accuracy (90%). Current RF sensing models to estimate moisture are destructive; they require samples to be constrained in containers. Our work is first to dispense with such impracticalities, and, without contact with the object, accurately measures non-uniform grape clusters in open space. We demonstrate that SING is superior to existing work in its ability to accurately measure the dielectric properties of non-uniform fruit objects and test this through both lab-based experimentation and preliminary outdoor vineyard tests. We also examine the transferability of SING’s approach to real-world scenarios.