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子粒两种不同样品类型的差异。102份绿豆核心种质样品为试验材料,采用近红外分析系统扫描了绿豆子粒和粉样品。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值为3.08~4.61。子粒样品最大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值为2.51~3.23。模型的稳健性采用外部验证法进行了评价。绿豆粉样品的平均差异为1.0%~1.8%,比子粒样品略低。结果表明绿豆粉3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。子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。 Near infrared spectroscopy was used to predict the contents of protein, amylase and amylose in the main nutritive qualities of Chinese mung bean and to compare the differences of two types of mung bean flour and grain samples. 102 samples of mung bean core collection were used as experimental materials, mung bean grains and powder samples were scanned by near infrared analysis system. After the preprocessing of spectral data, the least squares regression prediction and interaction verification model was constructed and the optimal calibration statistics were obtained. The mung bean flour samples had the highest R2 value and the lowest SECV value of 0.95 and 0.329 for protein content, 0.90 and 0.576 for starch content, 0.89 and 0.307 for amylose content, and 3.08-4.61 for RPD value. The maximum R2 and minimum SECV of the grain samples were 0.90 and 0.404 for protein content, 0.88 and 0.643 for starch content, 0.85 and 0.426 for amylose content, and RPD values ​​ranged from 2.51 to 3.23. The robustness of the model was evaluated using external validation. The average difference of mung bean flour samples was 1.0% ~ 1.8%, slightly lower than that of the seed particles. The results showed that the near-infrared prediction method of mung bean flour three components has the characteristics of fast and simple, and can be applied to the determination of mung bean quality. Seed samples also have the characteristics of non-destructive seed structure, maintaining seed vigor, which can be applied in breeding and germplasm resources innovation, but its accuracy needs to be further improved.