Dense Interpolation – Wavelength Shifting – Mean Spectra Subtraction Correction: A New Calibration Transfer Algorithm for Near-Infrared Spectroscopy Instruments on Crude Oil Fast Assay

Model transfer, also known as instrument standardization, constitutes a fundamental aspect of near-infrared (NIR) technology. Owing to disparities among spectrometers, spectra acquired from identical samples by different instruments may be erroneously classified as belonging to distinct samples. Consequently, spectral databases established on one spectrometer cannot be directly applied to other instruments, thus impeding the widespread adoption of NIR technology. Model transfer algorithms based on factor analysis often compromise spectral integrity, rendering the transferred spectra unsuitable for pattern recognition. To address this challenge and facilitate the universality of NIR spectral databases for crude oil across diverse instrument platforms, this study proposes a novel methodology, Dense Interpolation-Wavelength Shifting-Mean Spectra Subtraction Correction. This method integrates dense interpolation, wavelength alignment, and background compensation techniques. Through meticulous implementation, this approach effectively mitigates spectral deviations induced by instrument discrepancies. Employing the moving window correlation coefficient method, the transferred NIR spectra of crude oil undergo pattern recognition computations within the NIR spectral database established on the primary instrument. The results demonstrate a successful recognition rate exceeding 90 %, indicating the effectiveness of the proposed methodology in improving spectral consistency and enabling robust pattern recognition in NIR analysis of crude oil.