Improvement of calibration models using two successive orthogonal projection methods. Application to quantification of wine mannoproteins

The improvement of calibration in infrared spectrometry requires a reduction of the effect of interference, that is to say the chemical or physical factors that disturb the spectrum and reduce measurement accuracy. We propose a methodology based on two successive orthogonal projections (SOP) that makes it possible to take into account the influence of the other identified and characterised chemical compounds. A first OSC or EPO projection was based on the pure spectra of the compounds in the solution. A second projection (OSC method) was performed on the calibration set obtained after the first projection. These two projections minimise respectively the identified chemical influence factors and the unidentified influence factors. Application on wine mannoproteins in medium infrared spectrometry gave a significant improvement in the calibration obtained with two successive projections in comparison with a single orthogonal projection.