摘要
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 。
Using intact rapeseed as test samples, some methods for spectra pretreatment and different regression methods were compared to establish a better regression equation for oil content by near infrared reflectance spectroscopy(NIRS). Meanwhile, three kinds of sample cups with different capacity were used to screen the suitable regression equation for rapeseed breeding. It was indicated by the results that the third derivative was best for mathematical treatment of the spectra value than others. According to effects of calibration and validation, the suitable mathematics treatments were '2, 3, 3, 1', '2, 6, 6, 1' and '3, 3, 3, 1'for data collection. Scatter correction for spectra had significant effects on calibration except 'Detrend(D)' for oil content analysis, and 'Transformation of Standard Normal Variate (SNV)', 'SNV+D', 'Standard Multiplicative Scatter Corrections (Standard MSC)' or 'Inverse Multiplicative Scatter Corrections (Inverse MSC)' could be used in option. Regression method had obvious effect on calibration, and the best one was 'Modified Partial Least Square (MPLS)' method. The analysis effects of the sample with 8 gram size was better than those of the samples of 3 gramsize or '0.60~0.90' size. Although the later two had a bit low in prediction coefficient (RSQ), they had low Bias and SEP(C) and had high calibration parameters (R 2, 1 VR) with low SEC and SECV in calibration set. These NIRS equations of rapeseed could be used to screen the better plants with high oil content in the early generations or intermediate lines in rapeseed breeding.
出处
《作物学报》
CAS
CSCD
北大核心
2002年第3期421-425,共5页
Acta Agronomica Sinica
基金
教育部"高等学校骨干教师资助计划"
浙江省"15 1人才基金"项目资助
关键词
油菜籽
近红外反射光谱
含油量
回归方程
Rapeseed
Near infrared reflectance spectroscopy
Oil content
Regression equation