Application of iPLS-SPA variable selection in nondestructive testing of Spirulina powder

The feasibility of using visible and near infrared (Vis-NIR) spectroscopy was evaluated for nondestructive testing of Spirulina powders. Soft independent modeling of class analogy (SIMCA) was used to establish Vis-NIR spectral calibration model. A correct answer rate (CAR) of 93.33% for the discrimination of three varieties was obtained on full-spectrum. A hybrid variable selection algorithm based on interval partial least squares (iPLS) and successive projections algorithm (SPA) was proposed for the effective spectral variable selection. Five optimal effective variables were selected by that hybrid algorithm from 675 variables of full-spectrum. The CAR of 96.67% for the prediction set was obtained. Compared with the SPA based on the full-spectrum, visible spectra or NIR spectra, SPA based on could reduce the calculation time. The results show that it is possible for the nondestructive testing of Spirulina powders using Vis-NIR spectroscopy, and iPLS-SPA is an effective spectral variable selection algorithm.