论文部分内容阅读
随着土壤遥感科学的快速发展,光谱技术及其相关理论分析在农业生产与生态规划领域得到了广泛的应用。通过对光谱反射率进行一阶微分与倒数对数形式的变换,选取相关系数较大的波段构建基于多元逐步回归与偏最小二乘法的反演模型,得到如下基本结论:土壤光谱的一阶微分的变化,显著提高了土壤有机质含量与光谱间的敏感性;在多元逐步回归分析中,反射率一阶微分的多元回归模型的稳定性及预测精度最好;在偏最小二乘法回归分析中,反射率倒数对数模型的决定系数达到了最高的0.962,而总均方根误差为最低的1.082,其模型的稳定性及预算精度优于其他模型;总体上,偏最小二乘法回归模型的稳定性及预测精度优于多元回归模型,能够进一步满足研究区土壤有机质含量估算的实际需求。
With the rapid development of soil remote sensing science, spectroscopy and its related theoretical analysis have been widely used in agricultural production and ecological planning. Through the first-order differential and reciprocal logarithmic transformation of the spectral reflectance, the inversion model based on multivariate stepwise regression and partial least squares is selected to select the band with larger correlation coefficient, and the following basic conclusions are obtained: First-order differential of soil spectrum , The sensitivity and the sensitivity of the soil organic matter content to the spectrum were significantly increased. In the multiple stepwise regression analysis, the multiple regression model with the first order differential reflectance showed the best stability and prediction accuracy. In the partial least squares regression analysis, The coefficient of determination of reciprocal logarithm model reached the highest 0.962, while the total root mean square error was the lowest 1.082. The stability and budget of the model were better than other models. In general, the regression model of partial least squares regression was stable It is better than multivariate regression model in predicting the accuracy of the model and can meet the actual needs of soil organic matter content estimation in the study area.