摘要： 应用近红外光谱法(Near infrared spectroscopy,NIRS)和偏最小二乘法(Partial least squares,PLS)建立玉米粗蛋白质、粗脂肪和粗淀粉定量分析的近红外光谱数学模型,并对模型预测结果的准确性进行了评价。结果表明:近红外预测模型的内部交叉验证决定系数(R2cv)分别为:0.9778,0.9666和0.9927;交叉证实标准差(RMSECV)分别为:0.38,0.40和1.51;模型外部验证决定系数(Rv2al)分别为0.9391,0.9651和0.9875;外部验证标准差(RMSEP)为0.41,0.35和1.31。实际样品的常规分析结果得出玉米粗蛋白质、粗脂肪和粗淀粉的NIRS数学模型具有较高的预测准确性,可应用于玉米育种工作中的大批样品的品质分析。

关键词: 近红外光谱法, 玉米, 定标模型, 粗蛋白质, 粗脂肪, 粗淀粉

Abstract: NIR models of crude protein,crude fat and crude starch in maize were built using Partial Least Squares(PLS),and the predictive veracity of these models were evaluated.The R²_cv(coefficients of determinition for cross validation)for determining crude protein,fat and starch were 0.977 8,0.966 6 and 0.992 7 respectively,as the RMSECV were 0.38,0.40 and 1.51.The R²_cvl(coefficients of determination for validation) were 0.939 1,0.965 1 and 0.987 5.As RMSEP are 0.41,0.35 and 1.31.Then,these models were all used to predict practical maize samples.Comparing with the true values from chemical analysis,the NIR models could predict veraciously.This result combined with the calibration results suggested that the NIR models we built could be used to a mass of sample analysis in maize breeding.

Key words: Near infrared spectroscopy, Maize, Calibration n}del, Crude protein, Crude fat, Crude starch

中图分类号: 

• S513