A correction method for mitigating absorbance discrepancies between near-infrared spectrometers through the incorporation of blended carbon-titanium dioxide powder

In near-infrared spectroscopy analysis, ensuring the accurate transfer of models between different instruments relies on maintaining the accuracy of instrument wavelengths and absorbance. To mitigate absorbance drift at different wavelength points, this paper proposes a near-infrared spectroscopy point-by-point quadratic polynomial correction method based on carbon-titanium dioxide powder samples. The method establishes a quadratic polynomial relationship model for the absorbance of each wavelength point between the main instrument and slave instruments. The study utilized two S450 grating-based diffuse reflection near-infrared spectroscopy instruments, with one serving as the main instrument and the other as the slave instrument. The point-by-point quadratic polynomial was employed to correct wheat spectra collected by the slave instruments, and a crude protein content prediction model for wheat was established, comparing it with linear regression correction. After correction, the average Euclidean distance of wheat spectra decreased by 66.71%, from 0.0937 to 0.0321, and the average peak-valley Euclidean distance decreased by 72.28%, from 0.0203 to 0.0056. The standard deviation of the predicted results decreased by 90.69%, from 1.4372 to 0.1338. The correction effect of the method combined with traditional preprocessing methods was superior to using preprocessing methods alone. Overall, the near-infrared spectroscopy point-by-point quadratic polynomial correction method based on carbon-titanium dioxide powder samples significantly reduces spectral differences between different instruments, enhances spectral consistency, and diminishes prediction errors, achieving improved model sharing between instruments.