论文部分内容阅读
为建立近红外光谱技术测定荞麦蛋白质与淀粉含量的方法,本研究以217份荞麦样品为试验材料,采用最小二乘回归预测和交叉验证构建近红外预测模型。分析表明:前处理采用多元散射校正法(MSC),维数(Rank)分别为5和5,光谱区间6803.9~6094.2/cm所建立的荞麦蛋白质与淀粉含量模型的预测效果较好,其决定系数(R~2)分别为0.9481和0.9167,交叉验证均方根(RMSECV)分别为0.68和2.08,相对分析误差(RPD)分别为4.39和3.46,均大于3.0,外部验证相关系数均大于0.96。本试验所建立的蛋白质与淀粉含量近红外预测模型具有较高的准确度和稳健性,可用于荞麦品质的快速测定。
In order to establish a method for the determination of protein and starch content in buckwheat by near infrared spectroscopy, 217 buckwheat samples were used as experimental materials, and the prediction model of near-infrared was established by least square regression and cross-validation. The results showed that the pretreatment was based on multivariate scatter calibration (MSC) with Ranks of 5 and 5, respectively. The predicted results of the buckwheat protein content and starch content model were good with the spectral range of 6803.9 ~ 6094.2 / cm, and the coefficient of determination (R ~ 2) were 0.9481 and 0.9167, RMSECV was 0.68 and 2.08, respectively. The relative analytical errors (RPD) were 4.39 and 3.46, respectively, both of which were greater than 3.0. The correlation coefficients of external validation were all greater than 0.96. The protein and starch content near-infrared prediction model established in this experiment has high accuracy and robustness and can be used for the rapid determination of buckwheat quality.