基于近红外光谱技术的柑橘罐头品质测定方法的研究

为了快速检测柑橘罐头主要营养指标的含量,实验对柑橘罐头进行光谱采集,通过偏最小二乘(PLS)线性拟合建立关于柑橘罐头总糖、总酸、黄酮和维生素C的近红外定量分析模型。通过分析比较不同的光谱预处理方法对近红外光谱模型的影响,其中经归一化预处理后建立的总糖和黄酮模型最优,相关系数分别为0.9156和0.9147,交互验证均方根误差(RMSECV)为0.6305%和2.7028 mg/100 g;平滑预处理后总酸的相关系数为0.8921,RMSECV为0.0539%;以原始光谱数据建立的维生素C模型的相关系数为0.9129,RMSECV为1.4323 mg/100 g;预测集总糖、总酸、黄酮和维生素C的相关系数分别为0.8519、0.8106、0.8334和0.8425,预测均方根误差(RMSEP)为0.8969%、0.0638%、3.6055 mg/100 g和1.9732 mg/100 g。实验结果表明:应用近红外光谱技术测定柑橘罐头中总糖、总酸、黄酮和维生素C的含量是可行的。 

Study on the method of canned citrus quality measurement based on the Near-infrared spectroscopy

In order to quickly detect the contents of main nutrition indicators of canned citrus,the experiment was conducted to spectral acquisition of canned citrus,by Partial Least Squares(PLS) linear fit to establish Nearinfrared quantitative analysis model on canned citrus total sugar,acid,flavonoids and vitamin C. Analyzed and compared the effects of different spectral pretreatment methods on Near- infrared spectroscopy models, in which the normalized model of total sugar and flavonoids were optimal,correlation coefficients were 0.9156 and0.9147,root mean square error of cross-validation(RMSECV) were 0.6305% and 2.7028 mg/100 g. Smoothed model of total acid correlation coefficient was 0.8921,RMSECV was 0.0539%. Model of vitamin C based on raw spectral data whose correlation coefficient was 0.9129,RMSECV was 1.4323 mg/100 g. Through validation,prediction set of total sugar,total acid,flavonoids and vitamin C correlation coefficient were 0.8519,0.8106,0.8334 and 0.8425,predict root mean square error(RMSEP) were 0.8969%,0.0638%,3.6055 mg/100 g and1.9732 mg/100 g. The results showed that Near-infrared spectroscopy for rapid detection of canned citrus total sugar,total acid,flavonoids and vitamin C content was feasible.