β-胡萝卜素作为一种重要的营养元素, 其在果蔬活体内的测定方法通常必须对样品进行破坏, 并且耗时、 费力。 采用拉曼光谱技术预测活体枇杷果实内β-胡萝卜素含量, 以高效液相色谱法(HPLC)检测值作为参考值, 采用高斯平滑对光谱数据进行预处理, 同时使用多项式拟合（PF）的方法对光谱数据进行背景荧光扣除, 以标准β-胡萝卜素的拉曼图谱为基础, 选取了三个特征频移, 以特征频移和截取的光谱建立了MLR模型、 PLSR模型以及LS-SVM模型, 取得了比较好的预测精度。 其中LS-SVM模型的预测精度最高, 其预测相关系数R2p达到了0.91, 表明通过拉曼光谱对活体枇杷内β-胡萝卜素的含量检测是实际可行的。

In china, researches on Raman spectroscopy in terms of foodstuff mainly focus on carbohydrates, fatty acids, proteins and vitamins. Conventional methods for determining the carotenoids content require the extraction of the samples as well as other cleanup steps. In this work, Raman spectroscopy is applied to get the measured value form loquats with different mature stage which are compared with the reference value get from High Performance Liquid Chromatography （HPLC）, in order to find new, fast, and nondestructive calibration methods for quantification of β-carotene content in loquat fruits. Least Squares Support Vector Machine and Partial least squares data processing methods are used to analyze the Raman spectra while PLS model has a prediction quality with the correlation coefficient of 0.845; the root-mean-square error of 0.022 μg·g-1 and LS-SVM model has a better prediction quality with the correlation coefficient of 0.910 with the root-mean-square error of 0.058 μg·g-1.