近红外光谱变量选择及其在苹果可溶性固形物含量无损速测中的应用

Variable selection of near infrared spectrum and its application in the non-destructive rapid detection for the soluble solid content in apples

中文关键词:  近红外光谱  无信息变量消除  可溶性固形物含量  苹果

英文关键词:near-infrared spectroscopy  uninformative variable elimination  soluble solid content  apple

中文摘要:

      目的 采用近红外光谱技术, 筛选有效变量对苹果可溶性固形物含量进行无损快速检测。方法 以改进无变量信息消除算法为变量筛选方法, 采用多元线性回归算法建立校正模型, 采用外部盲样对模型进行预测准确度评价。结果 基于改进无信息变量消除算法, 筛选1391、1435、1521、1589 nm 4个关键波长作为变量, 其所建校正模型的测定系数为0.6823, 校正误差均方根为1.06, 交互验证测定系数为0.6780, 交互验证误差均方根为1.06。外部验证测定系数为0.6585, 预测误差均方根为1.07。经F检验, 预测模型的预测值与测定值之间具有显著相关性。结论 该方法基本能够满足苹果可溶性固形物含量无损快速检测的需求, 并可为水果可溶性固形物含量无损快速检测仪器的研制提供一定的技术参考。

英文摘要:

      Objective To determine the soluble solid content (SSC) of apple rapidly and nondestructively by selecting effective variables using near infrared (NIR) spectroscopy. Methods The improved uninformative variable elimination algorithm was used as variable selection method. The calibration model was developed by multivariate linear regression algorithm and the accuracy of the model was evaluated by external samples. Results The correlation coefficient of the calibration model (R2) developed by the 4 key wavelength variables (1391, 1435, 1521, 1589 nm) which selected by the improved uninformative variable elimination algorithm was 0.6823, the root means square error of calibration (RMSEC) was 1.06, the correlation coefficient of cross validation (R2cv) was 0.6780, and the root mean square error of cross validation (RMSECV) was 1.06. The correlation coefficient of the prediction model (R2P) was 0.6585, and the root mean square error of prediction (RMSEP) was 1.07. The F-test showed a significant correlation between the predicted and measured values of the predictive model. Conclusion This method can basically meet the demand of non-destructive and rapid detection of soluble solid content in apple, and provide technical reference for the research of non-destructive rapid detection instrument for SSC in fruits.