Development of a non-destructive detection method for the contents of total soluble solids and titratable acid in loquat using visible/near-infrared hyperspectral imaging

The flavor and market value of loquat fruit are significantly influenced by their total soluble solids (TSS) and titratable acid (TA) contents. This study employed hyperspectral imaging (HSI) technology to detect the TSS and TA contents of loquat fruit, aiming to evaluate the feasibility of this non-destructive technique for assessing postharvest quality. A total of 118 loquat samples were analyzed, with spectral data extracted from regions of interest in hyperspectral images. Outliers were removed using the Monte Carlo (MC) algorithm, and feature wavelengths were selected via the competitive adaptive reweighted sampling algorithm (CARS). Subsequently, the predictive performances of the two partial least squares regression (PLSR) models, the nonlinear iterative partial least squares (NIPALS) and the simple partial least squares (SIMPLS), were systematically evaluated and compared. For the TSS content, the SIMPLS model outperformed NIPALS, achieving a determination coefficient ( ) of 0.8955, a root mean square error of prediction ( RMSEP ) of 0.4200, and a ratio of prediction to deviation ( RPD ) of 3.14 when validated using an independent prediction set. For the TA content, the NIPALS model performed better, yielding an of 0.7841, an RMSEP of 0.0397, and an RPD of 2.19. It was therefore concluded that HSI technology combined with PLSR had the potential for use in non-destructive analysis of key postharvest quality indicators, the TSS and TA contents of loquat fruit. • Monte Carlo algorithm is used for outliers removing. • CARS algorithm is used for feature wavelengths selecting. • Develop and assess PLSR models to determine TSS and TA contents in loquat. • HSI techniques achieved high accuracy in detecting TSS and TA contents in loquat.