Spectral Multivariable Selection and Calibration in Visible-Shortwave Near-Infrared Spectroscopy for Non-Destructive Protein Assessment of Spirulina Microalga Powder

Visible and shortwave-near-infrared spectroscopy was used to predict protein content in Spirulina powder. Three variable selection algorithms of partial least square based uninformative variable elimination, partial least square based genetic algorithm, and successive projections algorithm were analyzed. Results showed that it was necessary to operate partial least square based uninformative variable elimination before successive projections algorithm. Successive projections algorithm performed better than partial least square based genetic algorithm, because it had a better result and fewer selected variables. The results showed that partial least square based uninformative variable elimination-successive projections algorithm was a good hybrid algorithm for the multivariable selection, and it was feasible for using visible and shortwave-near-infrared spectroscopy to non-destructively estimate protein content in Spirulina powder.