Research on Near-Infrared Spectroscopy Calibration Transfer Algorithm Based on Piecewise Direct Standardization and K-Nearest Neighbor

To address the issue of sharing qualitative models between different near-infrared spectrometers, this study explores the effectiveness of calibration transfer techniques. Two different types of near-infrared spectrometers were used as master and slave instruments to collect spectral data of samples. After preprocessing the spectral data using Savitzky-Golay (SG) smoothing and Multiplicative Scatter Correction (MSC), the study innovatively applied linear interpolation and Piecewise Direct Standardization (PDS) methods to achieve calibration Transfer. The calibration transfer effect was evaluated by calculating the Euclidean distance between the master and slave spectral data and the recognition rate of the K-Nearest Neighbor (KNN) classification model established on the master instrument. The results showed that the proposed calibration transfer method significantly improved the similarity between the master and slave spectral data, reducing the Euclidean distance from 8.6 to 0.2 and increasing the spectral recognition rate by 65.2%, approaching the recognition performance of the master instrument. This study effectively enables the sharing of qualitative recognition models between different types of near-infrared spectrometers using calibration transfer techniques, broadening the application scope of laboratory spectrometer qualitative models and providing new solutions for related fields with broad application prospects.