Abstract: This study aimed at improving the transfer of a calibration model for determining soluble solids content (SSC) in apples between two near-infrared (NIR) spectrometers. A total of 89 red Fuji apple samples were examined for SSC using an Abbe refractometer, and spectra were recorded using the two NIR spectrometers. First, a partial least squares regression (PLSR) model for predicting SSC was established on the master instrument. Then, the moving window correlation coefficient (MWCC) method was employed to analyze the lateral offset between the spectra from different instruments. Based on this analysis, the wavelet transform-piecewise direct standardization (WT-PDS) algorithm was applied to transfer the master model to the slave instrument, and its performance was compared with those of direct standardization (DS) and piecewise direct standardization (PDS). The results showed that compared with DS and PDS alone, wavelet transform (WT) preprocessing significantly improved the prediction performance for samples on the slave instrument. The WT-PDS algorithm achieved the most significant improvement; the relative prediction deviation (RPD) increased from 2.951 3 to 4.029 8, the root mean square error of prediction (RMSEP) decreased from 0.637 3 to 0.466 2, and the number of transfer samples declined from 40 to 20 compared with DS. Therefore, the WT-PDS algorithm effectively suppressed highfrequency noise and background interference in spectral signals, thereby improving the accuracy of model transfer and offering a more effective solution to reduce complex systematic differences between portable nearinfrared spectrometers.

Key words: near-infrared spectroscopy; model transfer; soluble solids content; moving window correlation coefficient; wavelet transform-piecewise direct standardization algorithm