The role of artificial intelligence in enhancing non-invasive quality monitoring in fresh food products supply chains: A comprehensive review.

Fresh food products are highly perishable and prone to quality deterioration throughout the supply chain, while conventional invasive monitoring methods suffer from low efficiency, destructiveness, and limited real-time capability. In recent years, non-invasive monitoring technologies (NIMT) including spectroscopy, imaging, sensing, acoustic, and molecular approaches have gained increasing attention due to their rapid and nondestructive advantages. The integration of artificial intelligence (AI) has further enhanced data interpretation, modeling, and decision-making, accelerating the intelligent transformation of quality monitoring. For instance, a meta-surface-based sub-terahertz "meta-sticker" enables layer-resolved spectral sensing between peel and pulp, offering a low-cost and scalable paradigm for non-invasive fruit ripeness grading. Meanwhile, the YOLO-Shrimp model developed for Litopenaeus vannamei achieved a 95.23% accuracy in visual freshness monitoring, demonstrating the power of AI in decoding complex imagery. This review systematically summarizes research progress in AI-assisted NIMT over the past five years, covering applications across pre-harvest, post-harvest, processing, cold-chain logistics, and retail stages. It highlights the role of AI models in feature extraction, quality prediction, grading, and traceability, and underscores AI's contributions to improving accuracy, real-time monitoring, and automation. However, challenges such as limited data quality, model transparency, algorithmic bias, and high implementation costs remain significant barriers to large-scale adoption. Future research should focus on multimodal data fusion, lightweight and edge AI models, and the integration of digital twins (DT) and the Internet of Things (IoT) to enable intelligent, traceable, and sustainable supply chain management from farm to table.