Development of a non-destructive method for wheat physico-chemical analysis by chemometric comparison of discrete light based near infrared and Fourier transform near infrared spectroscopy

Abstract

Rapid and non-destructive spectroscopic methods were developed using discrete light based near infrared (NIR) and Fourier transform near infrared (FTNIR) spectroscopy and compared for efficient determination of physico-chemical characteristics of wheat grain. The FTNIR spectra were analyzed using partial least squares regression with various preprocessing techniques. The best model for moisture, protein, ash, fat, thousand kernel weight and hardness with lowest RMSECV values 0.60, 0.17, 0.03, 0.02, 0.7, 1.2 and maximum correlation coefficient (R2) 0.97, 0.95,0.87,0.90,0.95 and 0.82 respectively were obtained. The discrete light based NIR spectral data were analyzed using multiple linear regression. The best model for moisture, protein, ash, fat, thousand kernel weight and hardness with lowest RMSECV values 0.94, 0.34, 0.04, 0.05, 1.09, 1.35 and maximum correlation coefficient (R2) 0.96, 0.90, 0.87, 0.75, 0.97 and 0.88 respectively were obtained. Comparing both the methods, FTNIR with lower relative error percentage was found to be useful for routine analysis in wheat processing industries.