Vis-NIR measurement of soluble solids in cherry and apricot by PLS regression and wavelength selection

Experimental results are presented on the use of partial least squares (PLS) regression and wavelength selection for the definition of models for visible-near-infrared (Vis-NIR) evaluation of soluble solids content in fruits. First, the relatively easy to deal with-but still not studied in the literature-case of cherry fruit is presented in detail. By using a very simple selection scheme, involving the subsampling of the spectral interval from 600 to 1100 nm with a fixed step, accurate models were found, consistently showing very favorable combinations of SEC and SEP values, in the 0.50 degrees Brix range for a total variation of about 15 degrees Brix. Apricot fruit represented a more difficult species, and wavelengths to be included in the calibration had to be individually selected for the best results. Nevertheless, parsimonious models could be found, including a total of 38 spectral lines and leading to SEP values at the 0.75 degrees Brix level.