Spectral pretreatment and wavelength selection for soluble solids content in gannan navel orange

Abstract

The feasibility of using visible and near-infrared (Vis-NIR) spectroscopy with the fiber optic technique for a rapid measurement of soluble solids content (SSC) in intact Gannan navel orange was investigated. The relationships between Vis-NIR spectra and laboratory measurements of SSC in Gannan navel orange were developed by using partial least squares (PLS) and multiple linear regression (MLR) methods, based on different spectral pretreatments and wavelength selection methods. In this study, the model based on multiplicative scatter correction (MSC) performed slightly better than those of other spectral pretreatments. In the comparison with the full spectrum model, the MLR method combined with sensitive wavelength selection and PLS method with spectral interval regions could obtain better calibration models. The SECV and r cv of the MLR model were 0.620Brix and 0.85, while those for iPLS model were 0.580Brix and 0.87. The same prediction results (SEP = 0.64, r p = 0.85, RSE% = 4.87%) showed that MLR method combined with sensitive wavelength selection and PLS method with spectral interval regions could be used for optimizing the calibration models. More importantly, MLR and iPLS may lead to simple future application. Copyright © 2010 American Scientific Publishers All rights reserved.