不同种皮颜色花生糖含量近红外模型的构建

花生籽仁中的糖含量是影响食味品质的重要指标，建立快速测定糖含量的方法可有效提高食用型花生的检测效率。样品外观颜色是影响近红外分析的重要因素之一，按样品外观颜色分类定标（校正）更有利于提高模型的预测性能。研究选择不同糖含量的花生种质332个，采用色差仪将花生种质按种皮颜色分成黑紫色、红色和粉色三大类。采用3,5-二硝基水杨酸法、蒽酮乙酸乙酯法、蔗糖酶法分别测定籽粒中的总糖、可溶性糖及蔗糖含量。总糖含量分别在6.42%～39.53%（黑紫色籽粒）、9.66%～39.71%（红色籽粒）和8.52%～38.84%（粉色籽粒）之间；可溶性糖含量分别在2.4%～14.32%（黑紫色籽粒）、2.94%～13.75%（红色籽粒）和2.19%～14.53%（粉色籽粒）之间；蔗糖含量分别在0.92%～7.53%（黑紫色籽粒）、1.05%～7.23%（红色籽粒）和0.95%～7.99%（粉色籽粒）之间，变异系数均在33%以上。采用瑞典波通DA7250型近红外分析仪（950～1 650 nm）采集籽粒的近红外光谱值，选用基于全波段的偏最小二乘回归法（PLSR），通过对比单一和复合预处理方法，对比模型的相关系数和误差确定最佳预测模型。分别建立了黑紫色、红色、粉色花生籽仁的总糖含量、可溶性糖含量和蔗糖含量的近红外光谱定标模型，共计9个模型，预测相关系数（R_c）在0.883～0.925之间，预测均方根误差（RMSEC）在0.370～1.988之间。对总糖含量所建立的模型中，粉色种皮花生的预测相关系数R_c可达0.925，RMSEC为1.705；对可溶性糖含量所建模型中，黑紫色种皮花生的预测相关系数R_c可达0.921，RMSEC为0.667；对蔗糖含量所建的模型中，粉色种皮花生的预测相关系数R_c可达0.914，RMSEC为0.435。并分别用15份种质进行外部验证，9个模型的预测相关系数R_p在0.892～0.967之间，预测均方根误差RMSEP在0.327～2.177之间。本研究建立的近红外光谱模型可同步、快速地检测花生籽粒中的多种糖含量，为高糖含量的鲜食花生育种提供了技术支持。

Construction of Near-Infrared Model of Peanut Sugar Content in Different Seed Coat Colors

The sugar content of peanut seeds is an important indicator that affects the quality of eating. Establishing a method for rapidly determining sugar content can effectively improve the detection efficiency of edible peanuts. The appearance color of the sample is one of the important factors that affect the near-infrared analysis. Classification and correction according to the appearance color of the sample are more conducive to improving the model’s the model’s predictive performance. Therefore in this study, 332 peanut germplasms with different sugar content were selected, and the peanut germplasms were divided into three categories: black-purple, red, and pink, according to the seed coat color using a colorimeter. The 3,5-dinitrosalicylic acid, anthrone colorimetric, and sucrose enzymatic methods were used to determine the total sugar, soluble, and sucrose content in seeds, respectively. The total sugar content was 6.42%～39.53% (blackpurple peanuts), 9.66%～39.71% (red peanuts), and 8.52～38.84 (pink peanuts). The soluble sugar content was 2.4%～14.32% (blackpurple peanuts), 2.94%～13.75% (red peanuts), 2.19%～14.53% (pink peanuts), and the sucrose content was 0.92%～7.53% (blackpurple peanuts), 1.05%～7.23% (red peanuts), 0.95%～7.99% (pink peanuts), the coefficient of variation was above 33%. The Perten DA7250 near-infrared analyzer (950～1 650 nm) was used to collect the near-infrared spectrum values of the seeds. The total 9 NIR spectroscopy calibration models about the total sugar content, soluble sugar content, and sucrose content of blackpurple, red, and pink peanut seeds were established respectively through the Partial Least Square Regression (PLSR) method based on the whole band, and the single or compound pretreatment methods, and comparing the correlation coefficient and error of the models. The correlation coefficient ofcorrection (R_c) was 0.883～0.925, and the root means standard error of calibration (RMSEC) was 0.370～1.988. In the model of total sugar content, the R_c of pink seed coat peanuts was 0.925, and the RMSEC was 1.705. In the model of soluble sugar content, the R_c of blackpurple seed coat peanuts was 0.921, and the RMSEC was 0.667. In the model of sucrose content, the R_c of blackpurple seed coat peanuts was 0.914, and the RMSEC was 0.435. External verification was carried out using 15 germplasms. The correlation coefficient of prediction (R_p) of the 9 models was 0.892～0.967, and the root means standard error of prediction (RMSEP) was 0.327～2.177. In this study, near-infrared models could predict the content of several sugars in peanut seeds simultantly and rapidly, and provide technical support for ediblepeanut breeding of high sugar content.