高油玉米群体油分、蛋白质和淀粉含量近红外分析模型的构建  

A Model for Determination of Kernel Oil, Protein and Starch Contents with High-oil Maize by Near Infrared Reflectance Spectroscopy

作　　者：张俊[1] 张义荣[1] 卢宝红[1] 宋秀芳[1] 宋同明[1] 李建生[1] 

机构地区：[1]中国农业大学国家玉米改良中心北京市作物遗传育种重点实验室,北京100094

出　　处：《玉米科学》2007年第3期62-66,共5页Journal of Maize Sciences

基　　金：国家自然科学基金项目(30571165)

摘　　要：以BHO高油玉米F2∶3家系为材料,应用主成分空间和傅里叶变换近红外光谱分析技术,采用偏最小二乘回归法(PLS),建立了测定高油玉米子粒的油分、蛋白质和淀粉含量的近红外校正模型。预处理分别采用一阶导数+矢量归一化、一阶导数+多元散射校正及直线相减等方法,主成分维数分别为5、9、9。验证分析表明,所建立的油分、蛋白质和淀粉含量的校正模型的校正和预测效果最好,其校正决定系数(R2cal)分别为0.950、0.973、0.976,交叉验证决定系数(R2cv)和外部验证决定系数(R2val)在0.918～0.948,各项误差(RMSEE、RMSECV、RMSEP)在0.305%～0.721%。结果表明,所建立的高油玉米完整子粒品质性状三成分模型的准确度和精确度均较高,可以满足高油玉米群体大量样品无损品质分析的需要。Using near infrared reflectance spectroscopy（NIRS） and partial least square（PLS） regression, a model for determination of oil, protein and starch contents in intact kernel was established with the high-oil materials of F2：3 family lines in maize. The results showed that the calibration models for oil, protein, and starch contents with the first derivative ＋ vector normalization, the first derivative ＋ multiplication scattering correction, and the straight line subtraction were the best. The coefficients of determination of the three models for calibration（R^2cal） were 0.950, 0.973 and 0.976 respectively while the R^2cv and R^2val were ranged from 0.918 to 0.948. The root mean square error of estimation, the root mean square error of cross validation, and the root mean square error of prediction（RMSEE, RMSECV, and RMSEP） for oil, protein, and starch contents varied from 0.305% to 0.721%. These results demonstrated that it was feasible to use NIRS as a rapid, accurate, and non-destructive technique to estimate oil, protein, and starch contents in intact kernel of the high-oil maize.

关 键 词：高油玉米 子粒品质 近红外光谱 校正模型 

分 类 号：S513.01[农业科学—作物学]