Rapid evaluation of the quality of chestnuts using near-infrared reflectance spectroscopy

Near-infrared (NIR) diffuse reflectance spectroscopy was used to evaluate the quality of fresh chestnuts, which can be affected by mildew, water, and levels of water-soluble sugars. The NIR spectra were determined and then modeling was performed including principal component analysis - discriminant analysis (PCA-DA), soft independent modeling of class analogy (SIMCA), linear discriminant analysis (LDA), and partial least squares (PLS) methods. LDA model was better than PCA-DA model for the discrimination of normal and mildewed chestnuts, and the accuracy rates of calibration and validation were 100% and 96.37%, respectively. Normal and mildewed chestnuts were easily distinguished by the SIMCA classification and showed only 4.7% overlap. A PLS model was established to determine the water and water-soluble sugars in chestnuts. The R² of calibration and validation were all higher than 0.9, while the root mean square errors (RMSE) were all lower than 0.05, indicating that the established models were successful.