Nondestructive firmness measurement of the multiple cultivars of pears by Vis-NIR spectroscopy coupled with multivariate calibration analysis and MC-UVE-SPA method

Abstract

The feasibility of using multi-cultivar model for non-invasive and accurate determination of firmness in different cultivar of pears was studied based on visible and near-infrared (Vis-NIR) spectrometric technique. A total of 330 samples were prepared for three cultivars of pears including “Cuiguan”, “Huanghua” and “Qingxiang”. Multi-cultivar model and three separate individual-cultivar models were first established and compared using full spectral variables. Multi-cultivar model did better than any one individual-cultivar model for firmness prediction of all pear cultivars. In order to eliminate useless variables and improve the signal/noise ratio, the pretreated full spectra were calculated by different informative variable selection methods. Combination (MC-UVE-SPA) of both Monte Carlo-uninformative variable elimination (MC-UVE) and successive projections algorithm (SPA) was more effective than single MC-UVE or SPA. Based on MC-UVE-SPA, seventeen effective variables were selected from full spectral 1344 variables for firmness analysis of pears. Linear partial least squares (PLS) and non-linear least squares-support vector machine (LS-SVM) models were developed by using effective variables and then were compared. MC-UVE-SPA-LS-SVM model was proved to be optimal in all developed models. Its correlation coefficient for prediction set (Rpre), root mean square error of prediction (RMSEP) and residual predictive deviation (RPD) were 0.94, 0.91 and 2.93 for “Cuiguan” pear, 0.93, 0.92 and 2.72 for “Huanghua” pear and 0.92, 0.96 and 2.55 for “Qingxiang” pear, respectively. The overall results indicated that MC-UVE-SPA was a powerful tool to select the effective variables, and MC-UVE-SPA-LS-SVM is simple and excellent for the determination of firmness of three cultivars of pears.