绿豆主要营养品质近红外预测模型的构建与验证

运用近红外法预测了中国绿豆主要营养品质蛋白质、淀粉和直链淀粉含量并比较绿豆粉和籽粒两种不同样品类型的差异。102份来自绿豆核心样品为试验材料，采用近红外分析系统扫描了绿豆籽粒和粉样品。光谱数据经预处理后，构建了最小二乘回归预测和交互验证模型并获得了最优校正统计参数。绿豆粉样品最大R2值和最小SECV值蛋白质含量为0.95和0.329，淀粉含量0.90和0.576，直链淀粉含量0.89和0.307；RPD值3.08至4.61。籽粒样品最大R2值和最小SECV值蛋白质含量为0.90和0.404，淀粉含量0.88和0.643，直链淀粉含量0.85和0.426；RPD值2.51至3.23。模型的稳健性采用外部验证法进行了评价。豆粉样品的平均差异1.0%～1.8%比籽粒样品略低。结果表明绿豆粉的三种组分的近红外预测方法具有快速和简单的特点，可应用于绿豆品质的测定。籽粒样品还具有无损种子结构，保持种子活力的特点，可在育种、种质资源创新等方面应用，但其准确度还有待进一步的改善。

The development of near-infrared spectroscopy (NIRS) prediction model for the quality components of flour and intact seed in mungbean

Near infrared reflectance spectroscopy was applied to determine the content of protein, starch and amylase in the mungbean grown in China. One hundred and two samples which both in flour and intact seeds were scanned in NIR Systems (MPA, Bruker, Germany) with Infra soft software OPUS 6.5 version. The partial least-squares regression method was developed with cross validation after processing data. The best calibration statistics were obtained by optimization. In flour the largest R2 value and the lowest SECV were for protein (0.95 and 0.329) for starch (0.90 and 0.576) and for amylase (0.89 and 0.307) and the residual prediction deviation (RPD) value were from 3.08 to 4.61, respectively. In intact seeds largest R2 value and the lowest SECV were for protein (0.90 and 0.404), for starch (0.88 and 0.643) and for amylose (0.85 and 0.426) and RPD value were from 2.51 to 3.23, respectively. The robustness of the model was evaluated for external validation. The mean differences, which were found 1.0% to 1.8% in flour, were slightly lower than the differences in intact seeds. NIRS prediction of the three compositions in the flour was applicable as a rapid and simple method. For intact seeds this method was non-destructive and could preserve its vigorous vitality and could be used in seeds conservation and germplasm resources innovation. But further work need to be done to improve the accuracy in the intact seeds.