An anthocyanin-curcumin colorimetric array with internal standard-guided feature switching for monitoring volatile alkaline nitrogen gases in meat freshness assessment.

Smartphone-based colorimetric sensor arrays using natural dyes are promising for on-site meat freshness assessment, but varying illumination and imaging angles compromise reliability. To address this, we developed an anthocyanin-curcumin array incorporating an α-Fe2O3 internal standard and an internal-standard-guided dynamic feature-space switching (IS-DFSS) strategy and validated its effectiveness under complex imaging conditions using ammonia as the representative analyte because it is the dominant component of total volatile basic nitrogen (TVB-N). Eight colorimetric films and the internal standard film were hydrophobically modified and assembled into a 3 × 3 array. The water contact angle increased from 63.3° to 112.9°, ensuring signal stability in high humidity. The sensing units responded strongly to TVB-N components ammonia, trimethylamine (TMA), and dimethylamine (DMA). For ammonia, Euclidean distance increased from 1.23 to 24.7 at 50 ppm and 13.7 to 65.7 at 500 ppm over 120 min, while non-alkaline volatiles (ethanol, hexanal) and non-volatile food matrix components (protein, glucose, fat) were negligible. In pork at 4 °C, array response correlated with TVB-N content (R2 = 0.987). The RGB values from 174 ammonia-responsive and 126 environment-interference images were averaged over three replicates to construct datasets, deriving HSV and Lab features. IS-DFSS with CatBoost reduced the mean absolute error to 15.85 ppm (45.46% reduction) and the mean false-signal error to 27.49 ppm (33.97% reduction) versus single RGB features, achieving R2 = 0.972. This method provides a robust, portable platform for monitoring TVB-N components, laying the foundation for reliable meat freshness assessment under complex imaging conditions.