基于可见/近红外透射光谱的番茄红素含量无损检测方法研究

Nondestructive Determination of Lycopene Content Based on Visible/Near Infrared Transmission Spectrum

针对番茄内外部结构特征,搭建了可见/近红外透射检测系统,利用完整番茄透射光谱信息,对番茄红素含量进行无损伤快速检测研究。采集的原始光谱曲线经去趋势(DT)、标准正态变量变换(SNV)、多元散射校正(MSC)、归一化(NOR)、一阶导数(FD)预处理后分别用偏最小二乘(PLS)进行建模分析。其中SNV预处理后的模型效果最好,校正集和验证集相关系数分别为0.9771和0.9504,校正集和验证集均方根误差为0.9711和1.0496 mg/kg。为进一步提高模型的精度和稳定性,采用无信息变量消除法(UVE)、连续投影算法(SPA)、竞争性自适应重加权算法(CARS)3种方法单独或联合处理(UVE-SPA,UVE-CARS),对全光谱进行变量优选。经UVE-CARS处理后番茄红素预测模型效果最好,其校正集和验证集相关系数分别提高至0.9830和0.9741,均方根误差分别降低至0.6919和0.7680 mg/kg。最后,选用25个番茄样品对所建立模型进行了外部验证,UVE-CARS-PLS模型的预测集相关系数为0.9812,预测集均方根误差为0.7071 mg/kg,平均相对误差为4.3%。而作为比较的PLS模型的预测集相关系数为0.951,均方根误差为1.0610 mg/kg,平均相对误差6.0%,相比于全光谱PLS模型,UVE-CARS可以很大程度地简化模型,提高模型精度,降低检测的误差限。结果表明,基于自行搭建的番茄可见/近红外透射检测系统结合光谱处理方法,可以实现对生鲜番茄中番茄红素含量的快速、无损检测,为番茄红素定量检测提供了新方法。

Lycopene is an important nutrient quality of tomato and has attracted people's attention in recent years.The traditional way of detecting lycopene damages tomatoes and takes a long time.So it is necessary to explore a method for nondestructive detection of lycopene.Due to the large differences in tomato composition,we chose transmission as a detection method and designed a visible/near infrared transmission spectrum detection system.Using this system,tomato transmission spectra were collected and then pretreated with Savitzky-Golay detrend(DT),standard normal variable transformation(SNV),muliplication scattering correction(MSC),normalize(NOR)and first derivative(FD).Finally,partial least squares model was established after processing.The model established by the SNV pre-processed spectrum had the best effect.The correlation coefficients between the calibration set and the verification set were 0.9771 and 0.9504,respectively.The root mean square errors of the calibration set and verification set were 0.9711 mg/kg and 1.0496 mg/kg.Considering that the original spectrum contained too many independent variables,uninformative variable elimination(UVE),successive projections algorithm(SPA)and competitive adaptive reweighted sampling(CARS)were used to optimize the variables.The three methods constituted five schemes(UVE,SPA,CARS,UVE-SPA,UVE-CARS).Among these five treatments,the lycopene prediction model was the best after UVE-CARS.The correlation coefficients of the calibration set and validation set increased to 0.9830 and 0.9741,respectively,and the root mean square errors decreased to 0.6919 mg/kg and 0.7680 mg/kg,respectively.In addition,25 tomato samples were used to externally verify the established model.The correlation coefficient of the prediction set of the UVE-CARS-PLS model was 0.9812,the root mean square error of the prediction set was 0.7071 mg/kg,and the average relative error was 4.32%.As a comparison,the correlation coefficient of the prediction set of the full-spectrum model was 0.951,the root mean square error was 1.0610 mg/kg,and the average relative error was 5.981%.The results showed that compared with the full-spectrum PLS model,UVE-CARS-PLS could greatly simplify the model,improve the model accuracy,and reduce the detection error limit.The result showed that the method with the combination of visible/near-infrared transmission spectroscopy and spectral processing methods could be used for the rapid and non-destructive detection of lycopene content.