Rapid discrimination of different degrees of stir-frying of wine-soaked Ligusticum chuanxiong based on E-eye, flash GC e-nose, and FT-NIR combined with multi-intelligent algorithms.

Ligusticum chuanxiong (LC) is a common medicinal and food-similar Chinese herb. After being processed with wine, its medicinal effect is further enhanced. However, the processing technology of LC has long suffered from problems such as inconsistent operation and difficulty in unified quality control. This study comprehensively utilized machine vision, e-eye, e-nose, and FT-NIR technologies, integrating multi-source information such as appearance, odor, and spectra, combined with an optimized classification model. The results show that the LBP features, the GLCM texture features, and the near-infrared spectral data processed by SNV + 2D, with variable importance projection values greater than 1, when used in the PSO-RF model, can achieve a 100% classification accuracy, effectively identifying different degrees of stir-frying LC. In conclusion, the integration of multi-source information and intelligent optimization algorithms provides a new approach for the rapid identification and quality control of medicinal Chinese herb LC.