Assessment of SSC and soluble sugar content of three pear cultivars during storage using dielectric method

This study aims to develop models for predicting soluble solids content (SSC) and soluble sugar content of three different pear cultivars. The samples' SSC and soluble sugar content, as well as the dielectric spectra in the frequency range of 20–4500 MHz were first collected, then the dielectric penetration depth was analyzed using one-way ANOVA method, and finally the accuracy of different models for predicting SSC and soluble sugar contents including partial least squares regression (PLSR), support vector machine (SVM), back propagation - artificial neural network (BP-ANN), and extreme learning machines (ELM) were compared. Results showed that the mean dielectric spectra of three pear cultivars presented similar trends. The dielectric penetration depth in the researched frequency range reached the maximum sugar content determination depth (intact pears: 20 mm; pulp: 17 mm). The correlations between glucose and dielectric spectra under specific frequency for 'Yulu' (determination coefficient: R2 = 0.88) and 'Korla' (R2 = 0.94) pears were the highest, while in terms of 'Xuehua' pear, the correlation was relatively lower (R2 ≤ 0.64). The Rp2 (R2 of prediction set) of the optimal sugar prediction model based on the dielectric spectra of intact pear and pear pulp were successively between 0.77–0.88 and 0.85–0.93. This research proved the feasibility of predicting sugar content in different pear cultivars using dielectric spectra. [Display omitted] •Sweetness is determined by sugar content.•Dielectric spectroscopy has the advantage of high penetration depth.•PLSR, SVM, and BP-ANN acquired satisfactory results for sugar content prediction.