论文部分内容阅读
目的利用可见/近红外反射光谱技术无损检测新鲜马铃薯茎块中抗性淀粉的含量。方法使用光谱仪获取新鲜马铃薯在345~1100 nm波段范围内的漫反射光谱;分别使用Savitzky–Golay(S-G)平滑处理、多元散射校正(MSC)法和一阶导数法(1st-D)对反射光谱进行预处理;对(S-G)反射光谱、MSC处理光谱和1st-D光谱使用逐步回归法判别法选择最优波长组合,建立多元线性回归模型,使用全交叉验证法验证模型。结果结果表明,可见/近红外反射光谱经过一阶导数处理后,确定的8个最优波长(370、569、576、866、868、886、922和963 nm)组合建立模型的校正和验证结果最好:模型的校正结果为相关系数R=0.996,标准差SEC=0.521%;模型交叉验证相关系数Rcv=0.982,验证标准差SECV=0.791%。结论可见/近红外反射光谱技术可以较好地预测新鲜马铃薯茎块的抗性淀粉含量,本研究可为可见近红外光谱技术在马铃薯功能成分的快速检测提供一定的技术基础。
Aim To evaluate the nondestructive detection of resistant starch in fresh potato stems by visible / near infrared reflectance spectroscopy. Methods The diffuse reflectance spectra of fresh potatoes in the 345-1100 nm band were obtained by using a spectrometer. The spectral reflectance spectra of Savoy-Golay (SG) smoothing, MSC and 1st-D spectra (SG) reflectance spectrum, MSC treatment spectrum and 1st-D spectroscopy were selected by the stepwise regression method to select the optimal wavelength combination, the establishment of multiple linear regression model, the use of cross-validation method to verify the model. The results show that the calibration and verification results of the model are obtained by combining the eight optimal wavelengths (370,569,576,866,868,886,922 and 963 nm) determined by the first-order derivative of the visible / near-infrared reflectance spectrum The best: the model calibration results for the correlation coefficient R = 0.996, standard deviation SEC = 0.521%; model cross-validation correlation coefficient Rcv = 0.982, validation standard deviation SECV = 0.791%. Conclusion Visible / near-infrared reflectance spectroscopy can predict the resistant starch content of fresh potato tubers. This study can provide a technical basis for the rapid detection of functional components of potato by visible near-infrared spectroscopy.