摘要
提高土壤全氮含量检测的精确性对于精准农业具有非常重要的意义。本文基于近红外光谱技术,对光谱数据预处理方法进行研究。以土壤全氮含量为检测对象,采用平滑+一阶导数、基线校正+归一化、一阶导数+归一化等方法对采集样品的光谱数据进行预处理方法干预,并运用PLS算法分别建立土壤全氮含量的预测模型,分析不同种预处理方法对于近红外光谱建模精度的影响。实验结果表明应用基线校正+归一化处理后建模得到的预测均方根误差RMSEP为0.025,决定系数R^2为0.98,预测精度最佳。
Increasing the accuracy of soil total nitrogen determination is of great importance for precision agriculture.Based on NIRS,the pretreatment method of spectral data is studied.Using soil total nitrogen as the test object,the spectral data of the collected samples were preprocessed using smooth+first derivative,baseline correction+ normalization and first derivative+normalization,and were modeled with PLS algorithm.The prediction model of soil total nitrogen content was used to analyze the effects of different pretreatment methods on the accuracy of NIRS modeling.The experimental results show that the RMSEP is0.025,R^2 is 0.98 after baseline correction+normalization,which has the best prediction accuracy.
出处
《光谱学与光谱分析》
SCIE
EI
CAS
CSCD
北大核心
2018年第S1期77-78,共2页
Spectroscopy and Spectral Analysis
基金
国家自然科学基金项目(41771357)资助
