高光谱成像技术结合化学计量学可视化花生中蛋白质含量分布

发布时间：2018-07-28 12:51

【摘要】：花生中蛋白质含量与分布能够显著影响花生制品品质。利用高光谱图像结合化学计量学研究可视化花生中蛋白质含量分布的可行性。从校正后的花生图像的感兴趣区域(region of interest,ROI)中提取光谱信息,通过传统化学方法测定蛋白质含量。比对了不同光谱预处理和回归算法,以二阶导数(the second derivative,2nd-der)为最佳的光谱预处理方法,偏最小二乘法(partial least squares,PLS)为最佳的回归算法。基于预处理后的光谱和花生蛋白质的化学值,建立全波长PLS模型,全波长模型具有良好的性能(校正集相关系数为0.91,校正集标准偏差0.86;预测集相关系数为0.86,预测集标准偏差为0.69)。利用回归系数法(regression coefficient,RC)从全波长模型中选择14个特征波长,建立2nd-der-RC-PLS特征波长模型,模型性能(校正集相关系数为0.86,校正集标准偏差1.03;预测集相关系数为0.80,预测集标准偏差为0.77)与全波长模型相当。采用2nd-der-RC-PLS算法将花生高光谱图像转变成蛋白质含量分布图。成对t检验判断凯氏定氮法与高光谱法无显著性差异。结果表明结合化学计量学的高光谱成像技术为测定花生中蛋白质含量分布提供了一种高效非破坏性方法。
[Abstract]:The content and distribution of protein in peanut could significantly affect the quality of peanut products. The feasibility of visualizing protein content distribution in peanut was studied by hyperspectral image and chemometrics. Spectral information was extracted from the region of interest (region of) of the corrected peanut image, and the protein content was determined by traditional chemical method. Different spectral preprocessing and regression algorithms were compared. The second derivative (the second derivation 2nd-der was the best spectral pretreatment method and the partial least square method (partial least squares was the best regression algorithm. Based on the pretreatment spectra and the chemical value of peanut protein, the full-wavelength PLS model is established. The full-wavelength model has good performance (correlation coefficient of correction set is 0.91, standard deviation of correction set is 0.86, correlation coefficient of prediction set is 0.86, standard deviation of prediction set is 0.69). By using the regression coefficient method (RC), 14 characteristic wavelengths were selected from the full-wavelength model, and the 2nd-der-RC-PLS characteristic wavelength model was established. The model performance (correlation coefficient of correction set is 0.86, standard deviation of correction set is 1.03, correlation coefficient of prediction set is 0.80, standard deviation of prediction set is 0.77) is equivalent to that of full-wavelength model. 2nd-der-RC-PLS algorithm was used to transform peanut hyperspectral image into protein content distribution map. There was no significant difference between Kjeldahl method and hyperspectral method by paired t test. The results show that the hyperspectral imaging combined with chemometrics provides an efficient and nondestructive method for the determination of protein content distribution in peanut.
【作者单位】：中国农业科学院农产品加工研究所农业部农产品加工综合性重点实验室;
【基金】：国家科技支持计划课题项目(2012BAD29B03) 中国农业科学院科技创新工程项目(CAAS-ASTIP-201X-IAPPST) 对发展中国家科技援助项目(KY201401005)资助
【分类号】：TS201.21;O657.3

【相似文献】