Surface Soil Moisture Estimation Using Visible Light Polarimetry

Abstract

Accurate soil moisture estimations are critical for agricultural and environmental monitoring. However, current methods rely on direct contact with the soil and can be invasive or limited in scalability. This work investigates a non-invasive optical approach using visible light polarimetry (VLP) to estimate soil moisture, expressed as volumetric water content (VWC). A polarization camera was used in a controlled lab setup to capture images of three soil types under polarized lighting. The degree and angle of linear polarization were extracted from these images and found to correlate with moisture changes. To account for variability across soil types, a convolutional neural network (CNN) was used to improve the accuracy of soil moisture sensing using VLP. This research contributes to the advancement of non-invasive soil moisture sensing techniques with potential deployment in real-world conditions to address the limitations of current methods.