Technology using near infrared spectroscopic and multivariate analysis to determine the soluble solids content of citrus fruit

Abstract

Several variable selection methods were investigated to select effective wavelengths (EWs) for determination of soluble solids content (SSC) of citrus fruits using NIR spectroscopy. The EWs selected by successive projection algorithm (SPA), stepwise multiple linear regression (SMLR), regression coefficient analysis (RCA) and x-loading weights analysis (x-LWA) were directly used for multiple linear regression (MLR). The variables selected by genetic algorithm (GA), uninformative variable elimination (UVE) and Monte Carlo UVE (MC-UVE) were used for partial least squares (PLS) regression. Finally the variables selected by GA, UVE and MC-UVE were further selected by SPA, SMLR, RCA and x-LWA, and corresponding MLR models were established. Results showed that the GA–SPA–MLR model yielded promising results, with rp and RMSEP being 0.893 and 0.436°Brix. The GA–SPA method was suggested to select EWs for MLR calibrations. Corresponding MLR function was provided and was supposed to be valuable for online SSC determination of citrus fruits.