摘要
以128份常用普通玉米自交系及杂交种的混合籽粒样品为材料,采用偏最小二乘(PLS)回归法,对近红外反射光谱(NIRS)测定玉米完整籽粒蛋白质、淀粉含量的可行性和方法进行了研究。结果表明,采用一阶导数+多元散射校正预处理、谱区为10000～4000cm-1和一阶导数+直线扣减预处理、谱区为9000～4000cm-1,分别建立的蛋白质、淀粉含量的校正模型,其校正和预测效果最佳。其校正决定系数(R2cal)均大于0.97,交叉验证和外部验证决定系数(R2cv、R2val)为0.92～0.95,各项误差(RMSEE、RMSECV和RMSEP)均小于1(0.3～0.7之间)。在玉米品质改良实践中,利用近红外反射光谱(NIRS)分析法,快速、准确、无损地测定完整玉米籽粒的蛋白质、淀粉含量是完全可行的。
Using 128 bulk- kernel samples of maize inbred lines and hybrids as materials, a study was conducted toinvestigate the feasibility and method of measuring protein and starch contents in intact seeds of maize by near infraredreflectance spectroscopy (NIRS). The chemometric method of partial least square (PLS) regression was used. The resultsindicated that the calibration models developed by the spectral data pretreatment of the first derivative + multivariatescattering correction within the spectral region of 10000-4000 cm-1, and the first derivative + straight line subtraction in9000-4000 cm-1 were the best for protein and starch, respectively. All these models yielded coefficients of determination ofcalibration (R2cal) above 0.97, while R2cv and R2val ranged from 0.92 to 0.95 for cross and external validation, respectively.However, the root of mean square errors for calibration, cross and external validation (RMSEE、RMSECV and RMSEP)were below 1 (ranged 0.30.7), respectively. It has been demonstrated that it is feasible to use NIRS as a rapid, accurate, andnone-destructive technique to predict protein and starch contents of whole kernel in maize quality improvement program.
出处
《中国农业科学》
CAS
CSCD
北大核心
2004年第5期630-633,共4页
Scientia Agricultura Sinica
