油菜籽芥酸和硫甙含量近红外反射光谱测定技术的优化设置

以完整甘蓝型油菜籽为样品,研究了不同光谱预处理和回归统计方法在用近红外反射光谱（NIRS）分析油菜籽中芥酸和硫甙的含量时,对建立回归方程的影响。结果表明,光谱预处理对校正结果影响较大,不同光谱数学处理以2阶导数处理较好,在各阶导数处理中存在着交叉现象,对于芥酸和硫甙都可找到一些较好的处理组合。对于芥酸,除趋势变换法（De-trending）散射处理外,其余各处理均有明显效果。但散射处理对硫甙分析效果不如芥酸明显。回归统计方法对建立芥酸和硫甙的回归方程影响最为有效,两者表现一致,其中以改良偏最小二乘法（MPLS）效果最好。对于芥酸和硫甙,采用“标准正态变量转换（SNV） 趋势变换法（De-trending）/2,4,4,1/MPLS”的组合建立回归方程效果较好,检验平均偏差（Bias）分别小至-0.405和0.315,而检验决定系数（RSQ）分别高至0.982和0.972。本研究采用整粒小样品（3g）来分析,效果较好,可直接用于育种早世代选择。

Calibration Optimization for Analysing Erucic Acid and Glucosinolate Content of Rapeseed by Near Infrared Reflectance Spectroscopy(NIRS)

Using intact kernel rapeseeds as test samples, various methods for spectra pretreatment and different regression methods were compared to establish an optimal regression equation for rapeseed erucic acid and glucosinolate content analysis by near infrared reflectance spectroscopy ( NIRS) . The results indicated that the second derivative was better for mathematical treatment of the spectra value than the first or third drivative, and the suitable Gap - Smooth - Smooth 2 combination was "2,4,4,1 "for data collection. Scatter correction for spectra had significant effects on calbration except De -trending (D)for erucic acid analysis.Standard normal variate (SNV)added D could be used in option,but for glucosinolate scatter correction methods had less significant effects than for erucic acid. Regression method had obvious effect on calibration,and the best one was Modified Partial Least Square (MPLS)method.The optimal calibration procedure could be"SNV D/2,4,4, 1/MPLS"for NIRS analysis of erucic acid and glucosinolate content of rapeseed with intact kernel sample. When using such calibration paramenters, the ideal regression equations were obtained with Bias for erucic acid and glucosinolate content as low as-0.405 and 0.315,and RSQ as high as 0.982 and 0.972, respectively. Better results were obtained by using intact kernel small sample (3g) .So the NIRS analysis could be used to screen the early generation lines in rapeseed breeding for quality impravement.