Assessment of Influence of Origin Variability on Robustness of Near Infrared Models for Soluble Solid Content of Apples / 分析化学

ABSTRACT

In order to improve the precision and robustness in determination of soluble solids content ( SSC) of ‘Fuji ’ apple by NIR spectroscopy and eliminate the effect of origin variability on the accuracy of NIR calibration models for the SSC, sample set partitioning based on joint x-y distances ( SPXY) was used to select representative subset from the apple samples of 4 different origins. As a comparison, partial least square ( PLS) was used to establish local origin and hybrid origin models for the prediction of SSC in apple. Competitive adaptive reweighted sampling ( CARS ) and successive projections algorithm ( SPA ) were implemented to select effective variables of the NIR spectroscopy of SSC of apple. The results indicated that the PLS model established based on the 4 origin apple samples performed better than local origin and other hybrid origin models. The model could be effectively simplified using 16 characteristic variables selected by CARS-SPA method from full-spectrum which had 3112 wavelengths. The correlation coefficient (Rp) and root mean square error of prediction (RMSEP) were 0. 978 and 0. 441 oBrix, respectively for SSC. It was found that the model developed by more samples of different origins combined with effective wavelengths showed good prediction ability for apple sample of unknown origin, which indicated that it could significantly reduce the origin effect on the robustness of NIR models for SSC of apple.