Near-infrared (NIR) and Raman spectroscopy have been extensively
utilized for analysis of various agricultural samples [1–5]
since these methods enable fast and non-destructive analysis
without requiring further sample pretreatment. The chemical
composition of agricultural samples is usually complex, severe
overlapping with individual bands of components make the obtained
spectral features indistinct. Therefore, for either quantitative
analysis or classification using spectra with complex features,
the adoption of a multivariate analysis such as principal component
analysis (PCA) and partial least squares (PLS) is now a routine
practice [6–12]. PCA and PLS describe linear variability in spectra
by linear projections of greatest variance from covariance matrix.
Although these methods have been successful for chemometric