摘要
采用偏最小二乘法（PLS）等方法建立了食用调和油中花生油含量定量分析的近红外光谱定标模型。采集食用调和油样品在4000cm^-1～10000cm^-1范围内的近红外漫反射光谱，光谱经一阶导数处理后，采用偏最小二乘法建立样品中花生油含量的定标模型，并用Leave—one—out内部交叉验证法对模型进行验证。模型相关系数为0．99961，校正均方根RMSEC为0．830％。比较不同光谱预处理方法对定标模型的影响，结果表明一阶导数Corr．coeff最好。采用不同的化学计量学方法建立的定标模型中以偏最小二乘回归法最理想。
Calibration models of quantitative analysis of the content of peanut oil in blended edible oil were built using partial-least-square （PLS） regression and other methods. Near infrared diffuse reflectance spectra of the calibration sample set were collected over 4 000 cm^-1 ~ 10 000 cm^-1 spectral region. Calibration model was established with PLS after the spectra were pretreated with first derivative and validated with leave-one-out cross validation method. The correlation coefficient was 0. 99961 and the RMSEC was 0. 830 % for the model. Among different spectra pretreatment methods, the first derivative was the best one. The eomparison of the calibration models built with different methods showed that the PLS was the optimal one.
出处
《激光生物学报》
CAS
CSCD
2007年第6期759-762,共4页
Acta Laser Biology Sinica
基金
国家质量技术监督检验检疫总局技术装备项目
浙江省教育厅项目(20050382)
关键词
食用调和油
花生油
近红外光谱
偏最小二乘法
定标模型
blended edible oil
peanut oil
near-infrared spectroscopy（NIRS）
partial least squares （PLS）
calibration model
