Raw spectral data for geographical origin traceability of Chrysanthemum morifolium (Gongju) based on HSI.

1.The central hypothesis of this study is that the "Dao-di" quality traits of Gongju (white petals, green receptacles, and intact forms) are deeply linked to its specific geographical origin, and that these differences can be captured through both biochemical analysis and hyperspectral imaging (HSI). We hypothesized that variations in ecological conditions and processing methods in She County, Anhui Province, create a unique "spectral fingerprint" and bioactive profile that distinguishes authentic Gongju from samples produced in other regions.
2.Data Characteristics and Gathering
The dataset consists of 1,440 data points derived from authentic and adulterated samples. The data were gathered using a dual-lens HSI system covering the VNIR (410–990 nm) and SWIR (950–2500 nm) ranges, totaling 396 wavelengths. To ensure data robustness, we flattened the samples to mitigate 3D scattering effects and utilized manually selected Regions of Interest (ROIs) to extract calibrated spectral cubes rather than point-based data.
3.Notable Findings and Interpretations
Our data shows significant reflectance differences, particularly in the 1500–2500 nm range, which we interpreted as vibrations from key bioactive constituents. Notable findings include:
①Optimal Discrimination: The Second Derivative (SD) combined with the PLS-DA model achieved a classification accuracy of >99%, effectively suppressing baseline drift and enhancing subtle spectral variances.
②Quantitative Precision: For predicting adulteration levels, the SVR model demonstrated superior performance (RPD > 6.0), indicating that the relationship between reflectance and chemical composition is better captured through nonlinear modeling.
③Chemical Correlation: Wavelengths identified by the MASS algorithm (e.g., 1400–1450 nm and 1900–1950 nm) correlate to the O-H combinations found in flavonoids and phenolic acids, providing a chemical basis for the models.
4. Data Interpretation and Usage
Others can interpret this data as a multidimensional map of herbal quality where spectral intensity reflects the concentration of medicinal compounds. By utilizing the feature wavelengths selected in this study, researchers can develop simplified, low-cost multispectral devices for rapid, on-site market surveillance and authenticity screening of high-end herbal products.