基于近红外光谱的脐橙产地溯源研究

Determination of geographical origin of navel orange by near infrared spectroscopy

为研究近红外光谱分析技术鉴别脐橙产地的可行性,该文采用江西、重庆和湖南3个产地脐橙样品1140～1170nm波段的近红外光谱经一阶导数(9点平滑)预处理,分别建立了簇类独立软模式法脐橙产地鉴别模型。在5%显著水平下,模型对3个产地训练集样品的识别率均为100%,拒绝率分别为85.7%、83.3%、100%;对验证集样品的识别率均为100%,拒绝率分别为100%、89.5%、100%,表明簇类独立软模式法模型基本能够判别脐橙产地。将江西、重庆和湖南3个产地的脐橙样品分别赋值0、1、?1,在全波段范围内建立原始光谱脐橙产地的偏最小二乘判别模型,其预测值与真实值的决定系数为0.973,校正标准差为0.110,预测标准差为0.159,模型对训练集和验证集样品的识别率达到100%。因此,应用近红外光谱分析技术可准确、快速地追溯脐橙产地来源。

In order to investigate the feasibility for identifying the geographical origin of navel orange by near infrared reflectance spectroscopy （NIRS）, three kinds of samples obtained from Jiangxi, Chongqing and Hunan province were tested in this paper. Three models of navel orange traceability were developed using the first derivative （9 points smoothing） of spectra at 1140-1170 nm combined with Soft Independent Modeling of Class Analogy （SIMCA）. Under 5% significance level, the identification rates of three models for the calibration set of samples were all 100%, and the rejection rates were 85.7%, 83.3% and 100%, respectively. The identification rates for the validation set of samples were all 100%, and the rejection rates were 100%, 89.5% and 100%, respectively. Give values 0, 1, -1 as category variables for origins of Jiangxi, Chongqing and Hunan, partial least squares-discrimination analysis models （PLS-DA） were established to determine geographical origins of navel orange. The best models were developed using raw spectra at 950-1650 nm when the number of principal components were 13. The results showed that the correlation between the predicted category variable and the measured category variable was significant with high correlation coefficient R2 （0.973）, low root mean square error of calibration RMSEC （0.110） and root mean square error of prediction RMSEP （0.159）. The identification rates were both 100% by PLS-DA model based on the calibration set and validation set of samples. These indicated that NIRS coupled with SIMCA and PLS-DA methods can be used for quickly and accurately discriminating geographical origin of navel orange samples.