Systematic Review on Artificial Intelligence‐Enhanced Hyperspectral Imaging for Food Processing: Applications, Technical Bottlenecks, and Future Perspectives

Abstract

Hyperspectral imaging (HSI) integrates spatial and spectral analysis for nondestructive food quality monitoring. While the integration of artificial intelligence (AI) has significantly advanced HSI analytics, prior reviews have predominantly focused on static, postharvest quality grading. This leaves a critical gap regarding dynamic physicochemical transformations during continuous food processing. To bridge this gap and provide a distinct engineering perspective, this systematic review comprehensively evaluates AI‐enhanced HSI applications across three fundamental mechanisms: mass transfer (drying), coupled biochemical‐mass transfer (pickling), and thermal processes (cooking). We critically synthesize the methodological transition from traditional chemometrics to advanced deep learning architectures, including convolutional neural networks (CNNs), graph neural networks (GNNs), and vision transformers. Crucially, we highlight their superior ability to autonomously extract spatiotemporal features from complex, deforming food matrices. Despite these algorithmic triumphs, large‐scale industrial translation remains hindered by prohibitive hardware costs, massive data latency, and poor model interpretability. To address these bottlenecks, we outline a concrete engineering roadmap prioritizing deep‐learning‐driven HSI reconstruction, edge‐AI deployment for real‐time actuation, explainable AI (XAI) to ensure regulatory trust, and direct programmable logic controller (PLC) integration. These multidisciplinary strategies aim to shift the paradigm from isolated offline inspection to closed‐loop automation, establishing AI‐enhanced HSI as the core perceptual engine for smart food manufacturing.