As a valuable Chinese herbal medicine， Panax notoginseng exhibits therapeutic efficacy and quality closely associated with its saponin content， which demonstrates significant geographical variations. To accurately authenticate the geographical origin and ensure medicinal quality， a novel method integrating terahertz precision spectroscopy with a convolutional neural network （CNN） algorithm was proposed. 40 Panax notoginseng samples from 4 regions in Yunnan Province， China—Honghe Autonomous Prefecture， Kunming， Qujing， and Wenshan Autonomous Prefecture—were analyzed using terahertz spectroscopy and high-performance liquid chromatography （HPLC）. A CNN model was constructed and trained based on the acquired spectral and chromatographic data to classify the geographical origins. Experimental results revealed that the terahertz spectroscopy combined with the CNN model achieved a classification accuracy of 92.5%， significantly outperforming the 82.5% accuracy attained by the HPLC-CNN model. This finding highlights the potential of terahertz spectroscopy in component analysis and geographical traceability of herbal medicines， providing a novel scientific approach for rapid， non-destructive， and precise identification of Chinese medicinal materials.