Reflectance Multispectral Imaging for Soil Composition Estimation and USDA Texture Classification
Researchers developed a cost-effective multispectral imaging system to predict soil composition and USDA texture classes using machine learning. By capturing thirteen spectral bands from field measurements, they compared three classification strategies—direct, regression, and indirect mapping—finding the direct method most accurate for texture classification. This offers a non-destructive, scalable tool for rapid soil characterization.