Design and Implementation of an Automated Protocol for Spectral Signatures Acquisition on Colombian Agricultural Soil Samples Into the Visible and Infrared Range

Pablo Andres Gomez Toloza, Henry Arguello, Ariolfo Camacho

Abstract: Visible and near-infrared (VIS–NIR) spectroscopy has emerged and developed in recent decades as a rapid, accurate, cost-effective, and non-destructive alternative method for soil analysis. This approach allows continuous soil monitoring and study to achieve sustainable agricultural management. However, spectral acquisition protocols are susceptible to spectral variations mainly induced by illumination factors, human interaction, or bias. This spectral variability produces additional problems related to machine learning algorithms or low-cost automated protocol implementation, to acquire spectral information that estimates Colombian agricultural soil properties to VIS-NIR spectroscopy. This work proposes an automatic acquisition protocol that minimizes the acquired data spectral variability in less acquisition time. Experiments show that the proposed automated spectral acquisition protocol on agricultural soil samples reduces the spectral variability up to 22%, using more than 1520 acquired spectral signatures at each experiment, obtaining higher quality in the resulting spectral data.