Optimization of NIR quantitative analysis model for wheat seed germination rate

The backward interval partial least squares (BiPLS) and the synergy interval partial least squares (SiPLS) were applied to select the characteristic spectral regions representing the germination rate of 84 wheat seeds and build the near infrared (NIR) quantitative analysis model of wheat seed germination rate.Results from comparison showed that the models built by two variable selection methods had better predictive ability than full-spectral partial least squares(PLS) model.The optimal model was obtained by SiPLS with the calibration and prediction correlation coefficient(R) at 0.902 and 0.967 respectively, and ratio of performance to standard deviate (RPD) at 3.75.Based on this, the physical chemistry significance of characteristic spectral regions was analyzed.The characteristic spectral of wheat seed germination rate contained characteristic peaks of water, protein, starch, fiber, which were the internal nutrients of the seed that influence the germination ability, thus explaining the mechanism of measuring wheat seed germination rate using NIR to a certain extent.