Spectroscopy-multimodal data fusion empowers smart food quality analysis: Challenges and prospects

Food quality assessment faces challenges in detecting complex samples, with single spectroscopic techniques limited in precision and applicability. Multimodal data fusion, by combining different spectroscopic techniques with non-spectroscopic data, significantly enhances the accuracy and comprehensiveness of detection. Low-, mid-, and high-level data fusion strategies offer substantial advantages in information integration and model optimization, addressing the shortcomings of single spectroscopic methods in complex food sample analysis. Multimodal data fusion has shown great potential in food adulteration detection, geographical traceability, and quality assessment. However, current technologies still face challenges such as data consistency, equipment integration, and model generalization. Future research will focus on lightweight equipment development, deep learning integration, standardization, and optimized data processing workflows to advance the development and application of food quality detection technologies.