Combination of visible and short-wave near infrared spectroscopy and chemometrics for determination harvest time of laver

Visible and short-wave near infrared (Vis-SwNIR) spectroscopy was employed to determination the harvest time of laver. In order to eliminate useless variables and improve accuracy, uninformative variable elimination (UVE) was proposed to eliminate the uninformative information from full-range spectra (FRS). Then, the retained variables were processed by Principal Component Analysis (PCA), and finally the 7 principal components (PCs) were inputted into least-square support vector machine (LS-SVM) to do the determination. A better result of 96.67% correct answer rate (CAR) of LS-SVM model was obtained. The obtained result was better than the LS-SVM model with no UVE processing. The result has indicated that UVE was an effective algorithm for elimination of uninformative variables. As a conclusion, Vis-SwNIR spectroscopy combined with chemometrics is a feasible way for determination of the different harvest time of laver.