基于高光谱的土壤有机质反演

Soil Organic Matter Inversion Based on Hyperspectral Data

摘要: 光谱预处理可以突显土壤光谱中的微弱信号，合适的光谱变换能够显著提升土壤有机质的预测精度，对土壤有机质的精准监测具有重要意义。本研究通过建立随机森林模型对土壤有机质进行估测，分析光谱变换对模型精度的提升。结果表明：一阶微分处理后的土壤有机质预测精度为0.93，均方根误差为0.911 g/kg，能够较好的预测土壤有机质含量。

Abstract: Spectral pretreatment can highlight the weak signal in the soil spectrum. The prediction accuracy can be improved by spectral transformation. It is of great significance for precision agriculture. In this study, the effect of spectral transformation is analyzed by predicting accuracy. The results show that: the prediction accuracy of the first-order differential treatment is 0.93; the root mean square error is 0.911 g/kg. It can better predict soil organic matter content.