Advances in hyperspectral imaging-based non-destructive assessment of tomato quality attributes

Tomato (Solanum lycopersicum L.), the world's second most economically significant horticultural crop, faces persistent quality assessment limitations due to conventional destructive methods and subjective inspections. Consequently, rapid non-destructive techniques for evaluating key quality attributes are essential to enhance postharvest management and market competitiveness. Hyperspectral imaging (HSI) addresses this need by acquiring spatially resolved spectral data across the biochemically critical 400–2500 nm range, capturing signatures of key attributes like lycopene and soluble solids. This technology demonstrates considerable potential for non-destructive agricultural assessment, enabling: high-accuracy maturity grading via pigment dynamics; ultra-early detection of asymptomatic diseases (e.g., latent fungal infections); quantitative nutritional component prediction; and reliable internal defect identification. These capabilities collectively enhance postharvest quality control. This review systematically evaluates advanced data preprocessing methods, feature extraction techniques, and deep learning models for their contributions to enhanced detection accuracy and system robustness. Nevertheless, challenges persist regarding hardware miniaturization, environmental resilience, model generalizability across diverse cultivars and growth environments, and real-time in-field implementation. Future efforts should prioritize developing integrated field-portable HSI platforms, multi-sensor fusion strategies, environmentally adaptive algorithms, and efficient AI architectures to bridge the gap between laboratory research and scalable industrial applications. Such advancements will critically underpin next-generation precision agriculture systems for robust tomato quality assurance and significant loss reduction.