Model for Non-destructive Quality Prediction of ‘Mihong’ Peach using Hyperspectral Imaging

This study aims to develop a non-destructive quality prediction model for 'Mihong' peach under open-field conditions using hyperspectral imaging.The primary objective is to identify wavelengths and vegetation indices that exhibit strong correlation with soluble solids content (SSC), titratable acidity (TA), and firmness in the image data captured by a hyperspectral camera.These indices were subsequently used to construct predictive models for SSC, TA, and firmness.Hyperspectral imaging data were collected from 120 'Mihong' peach samples during the 2023 and 2024 growing seasons.The data preprocessing methods (IQR, SNV, SG) were applied to improve the accuracy of the models, and predictive models were developed using partial least squares regression (PLSR), random forest (RF), and extreme gradient boosting (XGB) algorithms.Model performance was assessed using the coefficient of determination (R 2 ) and root mean square error (RMSE).