论文部分内容阅读
在用近红外反射光谱（ＮＩＲＳ）进行品质分析时，选择有代表性的样本进行校正是获得良好回归方程的重要步骤。运用ＩｎｆｒａｓｏｆｔＩｎｔｅｒｎａｔｉｏｎａｌ公司的ＣＥＮＴＥＲ和ＳＥＬＥＣＴ计算机程序，研究了稻米表观直链淀粉含量（ＡＡＣ）近红外测定建立回归方程时样品群体的界定与选择。试验共用１１０６份稻米精米粉样品为起始群体，根据光谱特征，有１６份样本因为标准化的Ｍａｈａｌａｎｏｂｉｓ距离（Ｈ距离）大于３．０而被剔除在校正和检验样本群体之外。在分别选用Ｈ距离≥０．０，０．４，０．６，０．８，１．０的样品作为校正群体时，样本数量分别为８９６，８４０，６４９，４３１和２７７个。对用其建立的ＡＡＣ回归方程，并用１个样本数为１９４个的独立群体对回归方程的效果进行检验。结果发现，用Ｈ≥１．０选定的２７７份样品群体和与用Ｈ≥０．０选定的８９６个样品群体建立的回归方程效果相似，校正决定系数Ｒ２和检验决定系数ＲＳＱ都高达０．９８以上，而标准误也几乎相似。从而表明，在建立稻米ＡＡＣ回归方程时，可用ＳＥＬＥＣＴ程序选择Ｈ≥１．０的代表性样本，以大大减少需实验测定的稻米样本的数量。
When performing quality analyzes with Near Infrared Reflectance Spectroscopy (NIRS), selecting a representative sample for correction is an important step in obtaining a good regression equation. Using the CENTER and SELECT computer programs of Infrasoft International Corporation, the definition and selection of sample population in the regression equation of rice apparent amylose content (AAC) near infrared spectroscopy was studied. Based on the spectral characteristics, 16 samples were excluded from the calibration and test sample population because the standardized Mahalanobis distance (H distance) was greater than 3.0. When the samples with H distance≥0.0,0.4,0.6,0.8,1.0 were selected as the calibration group respectively, the number of samples were 896,840,649,431 and 277 respectively. The AAC regression equation was established and the effect of the regression equation was tested by using an independent population of 194 samples. The results showed that the regression equation established by 277 samples with H≥1.0 and 896 samples with H≥0.0 were similar, the correction coefficient R2 and test determination coefficient RSQ were all 0 .98 or more, and the standard error is almost the same. Thus, when establishing AAC regression equations for rice, a SELECT program can be used to select representative samples with H≥1.0 to greatly reduce the number of rice samples to be experimentally determined.