可见/近红外漫透射光谱结合CARS变量优选预测脐橙可溶性固形物

可溶性固形物(SSC)是脐橙重要内部品质之一。采用QualitySpec型光谱仪在350～1000 nm波段范围采集脐橙的可见/近红外漫透射光谱，采用CARS(competitive adaptive reweighted sampling)变量选择方法筛选出与脐橙SSC相关的重要变量，并与无信息变量消除(UVE)及连续投影算法(SPA)比较。最后，对选择的38个重要波长变量应用偏最小二乘(PLS)回归建立脐橙SSC预测模型，并对未参与建模的75个样品进行预测。研究结果表明，CARS方法优于UVE及SPA变量选择方法，能有效地筛选出重要波长变量。CARS-PLS建立的SSC预测模型优于全光谱的PLS模型，其校正集及预测集的相关系数分别为0.948和0.917，均方根误差分别为0.347%和0.394%。因此，可见/近红外漫透射光谱结合CARS方法可以预测脐橙可溶性固形物，CARS变量选择方法能有效简化预测模型和提高模型的预测精度。

Determination of Soluble Solids Content in Navel Oranges by Vis/NIR Diffuse Transmission Spectra Combined with CARS Method

Soluble solids content (SSC) is one of important internal quality index for navel oranges. In the present study, visible/near infrared (Vis/NIR) diffuse transmission spectra of navel oranges were acquired using a QualitySpec spectrometer in the wavelength range of 350~1 000 nm, and CARS (competitive adaptive reweighted sampling) was used to select important variables related with SSC of navel oranges from spectra data, then was compared with other variable selection methods such as uninformative variables elimination (UVE) and successive projections algorithm (SPA). Finally, partial least squares (PLS) regression was used to develop calibration model for SSC of navel oranges using the 38 selected variables, and the calibration model was used to predict the SSC of 75 samples in the prediction set. The results indicate that CARS method is superior to other variable selection methods such as UVE and SPA, and can select the important variables for SSC efficiently. The calibration model of SSC developed by CARS-PLS is superior to that model developed by full-spectrum PLS, the correlation coefficient (r) and root mean square error (RMSE) in the calibration and prediction sets are 0.948, 0.347% and 0.917, 0.394%, respectively. So, Vis/NIR diffuse transmission spectra combined with CARS method is feasible to assess soluble solids content of navel oranges, and CARS method can simplify the prediction model and improve model prediction precision.