基于混合样品集的大豆种子蔗糖含量近红外光谱定标模型建立  

Establishment of NIRS Calibration Models for Sucrose Content in Soybean Based on the Combined Sample Set

作　　者：丛媛媛[1] 刘浩[2] 衣志刚[2] 王博[2] 刘念析 陈亮[2] CONG Yuan-yuan;LIU Hao;YI Zhi-gang;WANG Bo;LIU Nian-xi;CHEN Liang(Institute of Agricultural Economics and Information,Jilin Academy of Agricultural Sciences,Changchun 130033,China;Soybean Research Institute,Jilin Academy of Agricultural Sciences/National Engineering Research Center for Soybean,Changchun 130033,China)

机构地区：[1]吉林省农业科学院农业经济与信息研究所,吉林长春130033 [2]吉林省农业科学院大豆研究所/大豆国家工程研究中心,吉林长春130033

出　　处：《大豆科学》2020年第6期940-946,共7页Soybean Science

基　　金：吉林省科技发展计划项目(20200201136JC);中国博士后科学基金(2019M661234)。

摘　　要：为快速测定原料大豆的蔗糖指标,及时为调整豆制品生产工艺参数或原料配比提供数据支撑,以满足快速准确鉴定高蔗糖大豆育种新品系的需要,本研究建立基于混合样品集的大豆种子蔗糖含量近红外光谱定标模型。将天然和人工添加蔗糖的大豆样品结合起来作为混合参试样品集,先利用近红外光谱分析仪采集光谱数据,再通过酶比色法测定化学值,进行光谱数据预处理,采用偏最小二乘回归统计方法建立大豆蔗糖含量的近红外光谱定量分析模型,并利用未知蔗糖含量的大豆种子样品对模型的预测性能进行验证。以混合参试样品集建立的近红外光谱定标模型能够在较宽的蔗糖含量范围(37.83~139.35 g·kg^-1)内获得较为精确的预测结果,并将二阶导数联合多元散射校正作为光谱数据的最佳预处理方法。经外部验证,样品的化学测定值与定标模型预测值极显著正相关,且决定系数达0.943,预测均方差为3.17,相对误差为5.923%。所建立近红外光谱定标模型的预测能力较强、适用性较好,能够有效应用于大豆种子蔗糖含量的快速准确测量。This study established a near-infrared spectroscopy calibration model for sucrose content in soybean.It can provide data support for the rapid determination of sucrose in the raw materials of soybean,and then adjust the processing parameters or optimum material proportion of soy products.In addition,it can meet the needs of breeding high-sucrose soybean lines.In this study,natural and sucrose-added soybean samples were combined as the sample set.The spectral data was firstly collected using a near-infrared spectrometer.Then the near-infrared spectroscopy quantitative analysis model of soybean sucrose content was established by the determination of chemical value of enzyme colorimetry,spectral data preprocessing,and partial least square regression statistical method.Finally the prediction performance of the model was verified by using soybean seed samples with unknown sucrose content.In this study,the near-infrared spectroscopy calibration model based on the combined sample set could obtain more accurate prediction results in a wide range of sucrose content(37.83-139.35g·kg^-1).The second derivative combined with multivariate scattering correction was used as the best preprocessing method for spectral data.Through external verification,the chemical determination value of the sample showed a significant positive correlation with the prediction value of the calibration model,the determination coefficient was 0.943,the prediction mean square deviation was 3.17,and the relative error was 5.923%.The established near-infrared spectroscopy calibration model has strong predictive ability and good applicability,and could be effectively applied to the rapid and accurate measurement of sucrose content of soybean seeds.

关 键 词：大豆 品质 蔗糖含量 近红外光谱定标模型 

分 类 号：S565.1[农业科学—作物学]