Principal component analysis for the near-infrared spectroscopic imaging

Abstract

The reaction-diffusion phenomenon between HCl and NaOH in the microchannel was visualized. The wavelength used for visualization was selected from the distribution of regression coefficients by principal component regression analysis, and a regression model was created for simultaneous imaging of three components, HCl, NaOH, and NaCl. For multiple regression, the absorbance difference ΔA images at wavelengths of 1810 nm, 1900 nm, and 1920 nm were used. Visualization of the solution distribution was achieved for all components from the imaging images. However, in terms of concentration distribution, it was found that the accuracy was poor in quantitativeness. The accuracy of the regression can be improved by accurately measuring the original ΔA spectrum, indicating that the wavelength selection using principal component regression analysis may be significant.