论文部分内容阅读
随着高光谱遥感技术的快速发展,光谱技术已经在土壤理化性质、土壤养分等预测研究中得到了广泛应用。通过土壤高光谱反射率及其变形全氮含量的相关性,提取土壤光谱特征波段;采用多元回归和偏最小二乘回归法对全氮含量进行预测分析。结果表明:土壤光谱一阶微分显著提高了全氮与高光谱之间的敏感度;在多元逐步线性回归模型和偏最小二乘回归分析法建立的模型中,二者均能较好地进行预测,但在偏最小二乘模型中,反射率二阶微分的预测模型最高达到0.956,总均方根误差最低为0.045。其模型的稳定性和预测精度优于多元逐步线性回归所建立模型,可以更好地快速预测土壤全氮,为土壤质量的评价提供数据基础,也为研究土壤退化地区的预测与防治提供信息,对未来农业的发展具有重要意义。
With the rapid development of hyperspectral remote sensing technology, spectroscopy has been widely used in the prediction of soil physical and chemical properties and soil nutrients. The bands of soil spectral characteristics were extracted from the correlation between hyperspectral reflectance and total nitrogen content in soil. The total nitrogen content was predicted by multiple regression and partial least squares regression. The results show that the first-order differential of soil spectra significantly increases the sensitivity between total nitrogen and hyperspectral, and both of them can be predicted well in multivariate stepwise linear regression model and partial least squares regression analysis However, in the partial least squares model, the predictive model of second order differential reflectance reached a maximum of 0.956 with the lowest total root mean square error of 0.045. The model is more stable and predictive than the model developed by multivariate stepwise linear regression, which can predict soil total nitrogen better and provide a data basis for evaluating soil quality. It also provides information for studying the prediction and prevention of soil degradation, The future development of agriculture is of great significance.