论文部分内容阅读
以完整油菜籽为样品 ,通过光谱预处理和回归统计方法来优化油菜籽含油量的近红外反射光谱法所建立回归方程 ,同时采用三种不同用量的样品杯进行近红外扫描分析。结果表明 ,光谱预处理对较正结果影响较大 ,不同光谱数学处理以 3级导数处理较好 ,综合考虑定标组和检验组的效果 ,采用“2 ,3,3,1”、“2 ,6 ,6 ,1”和“3,3,3,1”的三种处理组合较好。除趋势变换法散射处理外 ,其余各处理均有明显效果 ,以标准正态变量转换、标准正态变量转换 +趋势变换法、标准乘性散射校正和逆向乘性散射校正处理效果最理想。回归统计方法对建立含油量的回归方程有很大的影响 ,其中以改良偏最小二乘法效果最好。总体上看 ,以 8g样品定标分析效果最好 ,3g样品分析略优于 0 .6～ 0 .8g样品分析 ,尽管这两种小样品检验组的预测决定系数 (RSQ)偏小 ,但两者预测值与实测值偏差的平均值 (Bias)、工作标准误 (SEP(C) )都很小 ;其回归方程的决定系数 (R2 ,1- VR)都较高 ,而相应的各项误差 (SEC,SECV)也较低 ,可直接用于品质育种中间材料和早世代材料的测定筛选
The regression equation established by near infrared reflectance spectroscopy with complete rapeseed as a sample, optimized by rapeseed oil content by spectral pretreatment and regression statistical method, and three different amount of sample cup for near infrared scanning analysis. The results show that the pretreatment has a great influence on the results of the correction. The results of different spectral mathods are better than the third derivative. Considering the effect of the calibration group and the test group, “2, 3, 3, 1”, “2 , 6, 6, 1 ”and“ 3,3,3,1 ”. In addition to the trend transformation method of scattering processing, the rest of the treatment has a significant effect, the standard normal variable conversion, standard normal-state transformation + trend transformation method, the standard multiplicative scatter correction and inverse multiplicative scatter correction effect is the best. The regression statistical method has a great influence on establishing the regression equation of oil content, of which the improved partial least squares method is the best. Overall, 8g sample calibration was the best, 3g sample analysis was slightly better than the 0.6 ~ 0.8g sample analysis. Although the predictive determination coefficients (RSQ) of the two small sample test groups were small, (Bias) and working standard error (SEP (C)) of predicted and measured values are very small. The determination coefficients (R2, 1-VR) of the regression equation are all high, and the corresponding errors (SEC, SECV) is also low, can be used directly for quality screening of breeding materials and early generations of materials screening