近红外光谱法测定棉粕中NDF和ADF的方法研究

Study On The Prediction Of The NDF And ADF In Cottonseed Meal By Near Infrared Spectroscopy

应用近红外光谱分析技术（Near infrared spectroscopy，NIRS）和偏最小二乘法（Partial least squares，PLS）建立棉粕中性洗涤纤维（NDF）、酸性洗涤纤维（ADF）定量分析模型，并对模型优化方法进行研究。模型的交叉验证决定系数（R^2ev）分别为0．94和0．95；交叉验证标准差（RMSECV）分别为0．668和0．441；模型外部验证决定系数（R^2val）分别为0．95和0．92；外部验证标准差（RMSEP）分别为0．81和0．44。结果表明：近红外光谱法可用于棉粕中NDF和ADF含量的快速测量。建模过程中选取有效的光谱区间可以提高模型的稳健性。

NIR models of NDF and ADF in cottonseed meal were built using Partial Least Squares （PLS）, and the methods of optimization on the models were studied. The R2cv were 0.94 and 0.95, as the RMSECV were 0. 668 and 0. 441; the R^2val were 0.95 and 0. 92, as the RMSEP were 0.81 and 0.44. The results show that NIRS could meet the needs of rapid evaluation. Choose an effective spectral range can improve the robustness of the model.