基于近红外光谱技术检测苹果气调贮藏期可溶性固形物含量

Detection of soluble solids content in apples during controlled atmosphere storage based on near-infrared spectroscopy

中文关键词:  贮藏苹果  可溶性固形物  可见/近红外光谱  主成分分析  预测模型

英文关键词:stored apple  soluble solids content  visible/near infrared spectra  principal component analysis  prediction model

中文摘要:

      目的 基于近红外光谱技术结合偏最小二乘(partial least square, PLS)法和最小二乘支持向量机回归(least square-support vector regression, LS-SVR)法建立苹果气调贮藏期可溶性固形物(soluble solids content, SSC)含量预测模型。方法 在分析了气调贮藏期苹果细胞结构和SSC变化的基础上, 采集了可见-近红外(visible-near infrared, Vis-NIR)波段和长波近红外(long wave near infrared, LWIR)波段下不同贮藏时间的苹果漫反射光谱信息, 利用主成分分析方法(principal component analysis, PCA)分析不同贮藏期苹果光谱信息分布特征, 使用Kennard-Stone (K-S)算法以3:1比例对样本集进行划分, 使用多元散射校正(multiplicative scatter correction, MSC)和卷积平滑(savitzky-golay, S-G)平滑对光谱进行预处理, 利用连续投影算法(successive projections algorithm, SPA)和竞争自适应重加权采样(competitive adaptive reweighted sampling, CARS)法对光谱进行特征波长提取, 并建立SSC预测模型。结果 在LWIR波段下, 经MSC+S-G预处理和CARS提取特征波长后建立的PLS模型取得了较好的预测精度, 模型相关系数为0.900, 均方根误差为0.478; 经MSC+S-G预处理和CARS提取特征波长后建立的LS-SVR模型取得了更好的预测精度, 模型相关系数为0.927, 均方根误差为0.507。结论 构建的基于可见/近红外光谱无损预测模型可实现对气调贮藏期苹果SSC的准确预测, 为高效贮藏技术提供了理论基础。

英文摘要:

      Objective To establish a prediction model for soluble solids content (SSC) during controlled atmosphere storage of apple based on near-infrared spectroscopy combined with partial least square (PLS) method and least square-support vector regression (LS-SVR) method. Methods On the basis of analyzing the changes in cell structure and SSC of apples during controlled atmosphere storage, diffuse reflectance spectra of apples at different storage times in the visible-near infrared (Vis-NIR) bands and long wave near infrared (LWIR) bands were collected. Principal component analysis (PCA) was used to analyze the distribution characteristics of apple spectral information in different storage periods. The sample set was partitioned in 3:1 ratio using the Kennard-Stone (K-S) algorithm, and the spectra were pretreated using multiplicative scatter correction (MSC) and savitzky-golay (S-G) smoothing. Successive projections algorithm (SPA) and competitive adaptive reweighted sampling (CARS) method were used to extract characteristic wavelengths from the spectrum, and build the SSC prediction model. Results In the LWIR band, the PLS model established after MSC+S-G preprocessing and CARS extraction of characteristic wavelengths achieved good prediction results, the correlation coefficient was 0.900, and the root mean square error was 0.478; the LS-SVR model established after MSC+S-G preprocessing and CARS extraction of characteristic wavelengths achieved better prediction results, the correlation coefficient was 0.927, and the root mean square error was 0.507. Conclusion The non-destructive prediction model based on visible/near-infrared spectroscopy can achieve accurate prediction of apple SSC during controlled atmosphere storage, and provide a theoretical basis for efficient storage technology.