Smart Egg Processing in the Digital Food Era, Part I: Advanced Sensing and Artificial Intelligence for Non-Destructive Egg Quality and Safety Assessment

Eggs are still frequently evaluated through periodic destructive tests, while high-throughput processing increasingly requires non-destructive measurements that can support timely decisions. This Part I critical review examines the sensing and prediction layer of smart egg processing, covering computer vision, hyperspectral and multispectral imaging, near-infrared, Raman and fluorescence spectroscopy, electronic-nose sensing, biosensors, and machine-learning methods used to interpret these signals. The evidence shows that no single modality provides complete coverage of external, internal, chemical, and microbiological attributes and that multimodal fusion is no longer merely conceptual: egg-specific studies have demonstrated the fusion of spectral and olfactory information and, more recently, image, shape, and weight features. Explainable AI has likewise entered egg research, including the SHAP-based interpretation of NIR and hyperspectral models. The more persistent limitation is the weakness and inconsistency of validation beyond the development dataset. Small or single-source datasets, incomplete reporting of preprocessing and feature-selection procedures, possible data leakage, class imbalance, limited calibration and uncertainty analysis, and scarce external or prospective validation constrain claims of industrial readiness. The review consequently reframes the central research gap as trustworthy generalization: demonstrating that sensing and AI remain accurate, interpretable, and decision-useful under the biological, environmental, and operational variability of real egg-processing systems.