Research on the Transferability of SSC Detection Models Between Different Instruments for Fresh Apricots

The determination of soluble solid content serves as a crucial parameter for assessing the internal quality of fresh apricots, thereby enhancing the precision of online detection and ensuring its significance in improving the accuracy of measuring soluble solid content in fresh apricots. To investigate spectral standardization methods for different NIR hyperspectral spectrometers, minimize spectral variations, and enable model sharing, we focused on soluble solids as the detection index and examined the spectra of fresh apricot samples collected by two NIR hyperspectral spectrometers. We employed direct standardization, slope/intercept correction, and single linear regression standardization algorithms to conduct our research on spectral standardization. The continuous projection algorithm is employed to enhance the model’s transmission capability by selecting the characteristic wavelength variable. The model integrates the single linear regression normalization algorithm with the preferred wavelength variable and can effectively be applied to spectral data collected using different instruments. Consequently, there is a significant improvement in the prediction set Rp2 from 0.1541 to 0.9405 and a remarkable decrease in RMSEP from 1.6161 to 0.3409, which facilitates model sharing across diverse instrument platforms.