Research on the grading method of Panax notoginseng slices based on machine vision

Precise grading is the foundation for improving the safety and consistency of Panax notoginseng in clinical applications. This study aims to propose a rapid, non-destructive quality grading method for Panax notoginseng based on machine vision and chemometrics. First, high-performance liquid chromatography (HPLC) was employed to determine the Panax Notoginseng Saponins (PNS) content of 143 samples. Based on hierarchical cluster analysis combined with the Elbow Rule, the samples were scientifically categorized into 3-Grade, 5-Grade, and 6-Grade standards. Five machine learning algorithms and six feature selection methods were compared to identify the optimal baseline model, and the Particle Swarm Optimization (PSO) algorithm was introduced to fine-tune the model's hyperparameters. The final model was evaluated using 10-fold cross-validation, and a bench test was conducted on 50 samples for grading verification. Comparative analysis identified the pearson correlation coefficient combined with CatBoost (COR-CatBoost) as the optimal baseline model across all grading schemes. After hyperparameter fine-tuning with PSO, the final COR-CatBoost-PSO model achieved average classification accuracies of 98.6% ± 0.5%, 88.2% ± 1.2%, and 84.5% ± 1.5% for the 3-Grade, 5-Grade, and 6-Grade standards, respectively, via 10-fold cross-validation. The bench test results showed a 100% classification accuracy of P. notoginseng, an average offset of the robotic arm of 1.2 mm, and a single grading process time of 0.8-1.5 s. The results verify the reliability and effectiveness of the proposed rapid, non-destructive quality grading method for Panax notoginseng, which can provide technical support for improving the safety and consistency of Panax notoginseng in clinical applications.