Abstract
The drying process is a critical step in determining the quality of safflower (Carthamus tinctorius L.). This study aims to systematically evaluate the effects of different drying methods on the chemical composition, color, morphology, odor, and microstructure of safflower dried petals, and to establish a comprehensive quality evaluation model for safflower based on machine learning. The results showed that drying methods significantly altered the chemical composition of safflower. Freeze-dried samples exhibited significantly higher levels of the active components hydroxysafflower yellow pigment A and anhydrosafflower yellow pigment B compared to other methods (p < 0.05), presenting a bright orange color and a mild odor. Microscopic structure and morphological analysis indicate that freeze-dried safflower effectively preserves its morphological characteristics, with a clear arrangement of cells and lower overall shrinkage. Based on nineteen quality parameters, nine quality evaluation models for safflower were constructed. The multiclassification decision forest model achieved a prediction accuracy of 89.1%. The importance analysis of quality parameters revealed that the B, G, and R color features in the RGB color mode are the most critical indicators for evaluating safflower quality. This study provides key basis for optimizing the drying process of safflower. The comprehensive evaluation model established provides a technical foundation for intelligent evaluation and standardized control of safflower quality, which is of significant practical value for improving safflower quality and promoting the standardized development of the industry.