Systematic data mixture

This study introduces a novel soil texture dataset designed to overcome geographic constraints and improve the generalization of classification models. Using the USDA soil classification triangle as a framework, the dataset is systematically generated by combining pure sand, silt, and clay in varying proportions to create diverse soil texture classes. The soil mixtures are captured using a multispectral sensor with seven bands, ensuring a rich representation of spectral information.