Multidimensional quality profiling of Gardenia Fructus: Rapid qualitative identification and quantitative prediction method based on near-infrared and electronic eye data fusion.

Gardeniae Fructus (GF), a key material for health products, teas, and natural pigments, currently lacks a methodology for comprehensive quality assessment. Therefore, an integrated quality control approach was developed combining color characteristics, spectral fingerprints, active ingredient content, and bioactivity of GF, which highlighted significant multi-dimensional variations among GF from different origins and processing methods. To achieve rapid traceability and prediction, E-eye and NIR technologies were integrated to acquire individual and fused signal features, employing three intelligent algorithms (Random Forest, Radial Basis Function, Particle Swarm Optimization Back Propagation Neural Network) for modeling and analysis. The processing methods of GF can be accurately classified using either color or spectral features alone, and the introduction of a feature-level fusion strategy enabled complete differentiation of GF from different origins, achieving an accuracy of 100%. In predicting the contents of geniposide, genipin gentiobioside, isochlorogenic acid C, and crocin I, as well as the bioactivity of GF against TNF-α and IL-6, optimized spectral features outperformed color features, while data fusion markedly improved the predictive effect, with RPD values all greater than 3 and R2 values close to 1. This study provides an innovative solution for the non-destructive, efficient evaluation and standardized control for GF quality.