Summary
This narrative review synthesises evidence on multimodal artificial intelligence systems that fuse heterogeneous sensor data—vision, spectroscopy, volatiles, biosensing and process telemetry—to enable real-time food safety and quality decisions from farm to retail. The authors detail data engineering practices for analysis-readiness, appraise fusion strategies from early/late to attention-based hybrids, and examine foundation model adaptation, edge deployment in industrial settings and regulatory alignment in EU, US and China frameworks. The review identifies persistent evidence gaps including few long-duration multisite deployments, limited public benchmarks for hyperspectral and e-nose fusion, and sparse cost–benefit analyses.
Regional applicability
The review addresses EU regulatory contexts directly and discusses alignment with EU standards alongside US and China frameworks. Findings on edge deployment, HACCP knowledge infusion and bias control in foundation models are transferable to United Kingdom food safety operations and compliance, though the UK's post-Brexit regulatory framework diverges from the EU standards discussed.
Key measures
Detection limits, response times, fusion architecture performance, accuracy improvements from multimodality versus unimodal baselines, edge deployment latency and reproducibility metrics
Outcomes reported
The review synthesises evidence on multimodal artificial intelligence systems that integrate vision, spectroscopy, volatiles, biosensing and process telemetry data to detect hazards, verify authenticity and predict freshness in real time across the food chain. It appraises fusion strategies, foundation model adaptation, edge deployment constraints and regulatory alignment across EU, US and China contexts.
Topic tags
Dig deeper with Pulse AI.
Ask about this record, its theme or its relevance to UK farming and policy. Pulse AI uses selected catalogue evidence and cites the sources it draws on.