近红外反射光谱测定玉米完整子粒蛋白质和淀粉含量的校正模型

Calibration Model of Determining Protein and Starch Contents of Whole Maize Kernel by Near Infrared Reflectance Spectroscopy

以玉米完整子粒为实验材料,采用偏最小二乘回归法建立近红外反射光谱测定其蛋白质和淀粉含量的数学模型,光谱预处理结果表明采用矢量归一化和一阶导数+多元散射校正分别建立蛋白质含量和淀粉含量的校正模型效果最佳,外部验证结果证明校正后模型预测结果与化学值之间的相关系数分别达到0.946 9和0.924 0。

Protein and starch contents of whole maize kernel were determined by near infrared reflectance spectroscopy（NIRS）. Math model was established by partial least square regression. Result of spectrum pretreatment indicated that the calibration models developed by vector normalization and first order derivative with multivariate scattering correction were the best for protein and starch, respectively. Exterior validation suggested that the correlations between the results of predicted NIRS by calibration and those obtained from chemical methods of protein and starch content were improved. The correlation coefficient is 0.9469 and 0.9240 respectively.