Systematic Dataset Generation for Soil Texture Classification Based on the USDA Soil Classification Triangle

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. This self-generated dataset enables the development and evaluation of advanced classification techniques, offering a standardized and comprehensive resource for soil texture studies. By addressing the limitations of existing datasets, this work provides a robust foundation for advancing soil texture classification research across diverse fields.本研究提出一款全新的土壤质地数据集，旨在突破地理限制，提升分类模型的泛化能力。该数据集以美国农业部（USDA）土壤分类三角为框架，通过按不同比例混合纯砂、粉粒与黏粒，系统生成多样化的土壤质地类别。研究人员采用七波段多光谱传感器对土壤混合物开展光谱数据采集，确保光谱信息得到充分且全面的表征。该自建数据集可用于开发与评估先进分类技术，为土壤质地相关研究提供标准化且全面的资源支撑。本研究通过解决现有数据集的局限性，为跨多领域推进土壤质地分类研究奠定了坚实基础。