Rapid identification and quantitative analysis of Ejiao beads at different processing degrees based on machine vision...

Traditional processing of Asini Corii Colla (Ejiao) involves stir-frying with clam powder, during which the duration of heating plays a critical role in determining product quality. This study systematically investigated the dynamic changes in characteristic polypeptides during the processing of Ejiao beads and proposed an integrated quality evaluation strategy combining machine vision with Fourier-transform near-infrared (FT-NIR) spectroscopy. The results demonstrated that appropriate spectral preprocessing was critical for classification performance, with BPNN achieving the most accurate and stable discrimination of processing degrees when combined with MSC or SNV + 1D, outperforming SVM and RF models. For quantitative analysis, the PLSR model for predicting the total content of donkey-derived polypeptides exhibited excellent performance (R²c = 0.9942, R²p = 0.9624), indicating strong linearity and high predictive accuracy. Overall, FT-NIR combined with chemometric modeling allowed precise identification of processing stages and accurate quantification of key polypeptide, providing an efficient and non-destructive method for quality control and standardization of animal-derived traditional medicines.