Evaluation of common pre-processing approaches for visible (VIS) and shortwave near infrared (SWNIR) spectroscopy in soluble solids content (SSC) assessment

The number of visible (VIS) and shortwave near infrared (SWNIR) spectroscopic applications in fruit internal quality has grown rapidly in the last decade. Despite this widespread application, pre-processed spectral data used is often not well understood. The aims of this paper are (i) to compare the use of SWNIR and VIS–SWNIR spectral data, (ii) to investigate the effect of different Savitzky–Golay (SG) derivatives (i.e. zero order, first order and second order derivatives) with different filter length, and (iii) to evaluate the use of log (1/R) transformation in the soluble solids content (SSC) assessment via Monte Carlo cross-validation (MCCV). Findings indicate that a parsimonious principal component regression (PCR) with four principal components achieved the best accuracy (i.e. root mean square error of cross-validation (RMSECV) = 0.81 °Brix and rcv = 0.75) when (i) visible spectrum was excluded, (ii) second order SG derivative with the optimal filter length was used, and (iii) the log (1/R) transformation was avoided.