土壤粒径的光谱响应特性研究

Study on Spectral Response Characteristics of Soil Particle Size

摘要:

以实验室制备的5个不同粒径水平的土壤样本和室内高光谱数据为基础，通过对光谱数据进行重采样、数学变换等预处理并进行单因素方差分析、相关性分析和回归分析，探讨土壤粒径的高光谱特性，建立了光谱数据预测土壤粒径的校正模型。结果表明，土壤粒径对反射光谱有显著的影响，波长越长影响越大；在全波段范围内土壤粒径和光谱数据都呈负相关关系，对原始光谱数据进行微分变换能增加其与土壤粒径的相关性；以反射率一阶微分建立的回归模型为反演土壤粒径的最佳模型，其建模决定系数R2c、预测决定系数R2p、预测相对偏差RPD分别为0.666，0.653，2.043，预测均方根误差RMSE为0.175。

Abstract:

Based on five soil samples with different particle sizes prepared in laboratory and indoor hyperspectral data, the hyperspectral characteristics of soil particle sizes have been discussed by pretreatment of spectral data, such as resampling, mathematical transformation, and one-way ANOVA, correlation analysis and regression analysis. Hyperspectral Characteristics of soil particle size has been studied, and a calibration model for predicting soil particle size based on spectral data has been established. It is showed that soil particle size has a significant effect on reflectance spectra, and the longer the wavelength is, the greater the effect is. Soil particle size and spectral data have negatively correlated in the whole band. Differential transformation of original spectral data could increase the correlation between soil particle size and soil particle size. The regression model based on first-order differential of reflectance has been inverted. The decision coefficient of modeling R2c, predictive determinant coefficient R2p and prediction relative deviation are 0.666, 0.653 and 2.043 respectively, and Prediction root mean square error RMSE is 0.175.