基于主成分回归的苹果可溶性固形物含量预测模型

为了更好地预测苹果的可溶性固形物含量（SSC），试验采用反射式光谱采集系统获取采后“富士”苹果的光谱反射率。分析了3种光谱预处理方法（标准正态变换、多元散射校正以及二阶导数）对预测模型的影响；利用主成分分析方法对预处理后的光谱数据进行降维，并基于选取的特征变量建立预测苹果SSC的回归模型。结果表明：采用主成分分析方法从全光谱的1 024个波长中选取了前23个主成分得分作为特征变量；基于特征变量建立的回归预测模型具有较好的预测能力，其预测集相关系数RP=0.908，均方根误差RMSEP=0.499。这表明采用光谱技术结合主成分回归预测苹果SSC是可行的。

Prediction Model for Soluble Solids Content of Apples Based on Principal Component Regression

In order to predict soluble solids content of apples, spectra acquisition system was used to collect the spectral reflectivity of postharvest apples. Then the influences of three spectral pretreatment methods including standard normal variation, multi-scatter calibration and second derivative on the prediction model were analyzed. And principal component analysis was used to conduct data mining from preprocessing reflectance spectra. And regression model was established based on selected characteristic variables for predicting SSC of apples. The results showed that, the first 23 principal components were selected as the characteristic variables by principal component analysis from 1 024 wavelengths. The regression model based on selected characteristic variables had the best prediction ability(RP=0.908, RMSEP=0.499). Therefore, it is feasible to predict SSC of apples by spectral technology combined with principal component regression.