应用傅里叶变换近红外光谱技术实现了鲜辣椒中可溶性固形物(SSC)和维生素C(Vc)含量的快速无损检测。分别采用7种预处理方法对原始光谱进行处理后, 建立了SSC和Vc预测的偏最小二乘法(PLS)模型。将利用最小二乘法(PLS)提取的主成分(PC)和蒙特卡罗无信息变量消除法 (MC-UVE)提取的有效波长作为最小二乘支持向量机 (LS-SVM)的输入变量, 分别建立了PC-LS-SVM 和MC-UVE-LS-SVM模型, 并与MC-UVE-PLS模型进行了比较。采用优化后的模型对27个预测集未知样品进行了预测。结果表明, 对鲜辣椒中SSC含量预测最优的为MC-UVE-PLS模型, 其预测集相关系数(rp)为0.971, 预测集均方根误差(RMSEP)为0.382 °Brix;对鲜辣椒中Vc含量预测最优的为MC-UVE-LS-SVM模型, 其rp为0.899, RMSEP为21.022 mg/100 g。研究结果表明: 鲜辣椒中SSC和Vc的含量与近红外光谱具有显著的相关性。

Fourier transform Near-infrared(NIR) spectroscopy was applied to the fast and nondestructive determination of Soluble Solid Contents (SSC) and Vitamin C (Vc) contents in fresh peppers. Seven kinds of pretreatment methods were used to the original spectral processing and the predicted Partial Least Square (PLS) models of SSC and Vc were established. The Principal Components (PC) selected by PLS and effective wavelengths selected by Monte Carlo Uninformative Variable Elimination (MC-UVE) method were used as the inputs of Least Square Support Vector Machine (LS-SVM), the PC-LS-SVM and MC-UVE-LS-SVM models were developed, and they were compared with the MC-UVE-PLS models. Twenty-seven unknown samples were predicted using the optimized models. The results show that MC-UVE-PLS model obtains the best result for SSC prediction with a correlation coefficient of prediction (rp) of 0.971 and Root Mean Square Error of Prediction (RMSEP) of 0.382 °Brix. The MC-UVE-LS-SVM model obtains the best result for Vc content prediction with rp of 0.899 and RMSEP of 21.022 mg/100 g. The research results indicate that SSC and Vc contents in fresh peppers have a significant correlation with the NIR spectroscopy.