Smart packaging is evolving to enhance food safety via real-time monitoring. This study developed electrospun nanofiber films composed of polyvinyl alcohol, polycaprolactone, and quaternary ammonium salt chitosan, loaded with shikonin and curcumin. Following shikonin and curcumin loading, nanofiber diameters increased from 100.2 to 229.0 nm while maintaining a uniform morphology. These nanocomposite films exhibited enhanced hydrophobicity, water vapor barrier properties, and mechanical strength. Incorporating 2% shikonin/ curcumin significantly improved antibacterial and antioxidant activities. Both the pigments solution and nanofibrous films demonstrated consistent pH-responsive behavior, transitioning from yellow at pH 2 to purple at pH 12, with a particularly pronounced shift under alkaline conditions. Practical application tests showed that the nanofiber films effectively monitored beef freshness over 10 days across different temperatures through distinct color changes. This research demonstrates the feasibility of utilizing dual-pigment electrospun nanofiber films as high-performance, active-smart materials for the food industry to ensure quality and safety.

Non-uniform heating is a major challenge in microwave food processing. To address this, we developed a biodegradable, microwave-active composite film by incorporating carboxyl-modified multi-walled carbon nanotubes (MWCNT-COOH) into a chitosan (CS) matrix. MWCNT-COOH markedly improved dielectric properties and formed a homogeneous, well-dispersed network within CS, enhancing interfacial compatibility and thermal conductivity. The optimized CS/MWCNT-COOH-0.9 film acted as an efficient microwave-absorbing and heat-conducting layer. When used as a film on model foods (rice balls) under 700 W microwave irradiation, it substantially improved heating uniformity, as evidenced by infrared thermography. The film reduced core-to-surface temperature gradients and eliminated hotspots and cold spots by converting microwave energy into heat via dielectric loss and Joule heating, followed by redistribution through thermal conduction. This work provides a practical packaging solution to enhance microwave heating for ready-to-eat foods and advanced processing technologies, with the final packaging exhibiting environmentally friendly characteristics.

Transportation-induced mortality from crowding, hypoxia, and handling stress costs aquaculture billions annually. Although anesthetic protocols mitigate these stressors, the anesthetic recovery phase remains critical for timely interventions. Invasive bioimpedance sensors introduce infection risks, computer vision fails in turbid water. Acoustic techniques demonstrate advantages yet face inadequate feature representation and severe dataset scarcity. To address these interdependent challenges confronting acoustic monitoring, this work presents a novel framework combining generative artificial intelligence (GAI) synthesis with multi-domain fusion for recovery stage discrimination. Specifically, Domain-Specific Denoising Diffusion Probabilistic Models (DSDDPM), synthesize biologically-authentic features across Zero-Crossing Rate (ZCR), Gammatone, Phase, and Wavelet domains, addressing data scarcity while preserving physiological validity. Subsequently, an adaptive progressive fusion strategy organizes features into periodicity and intensity groups, integrating complementary respiratory patterns through attention mechanisms. The lightweight MambaVision-MDF (Multi-Domain Fusion) architecture then processes the augmented fused features through dual-path temporal-spectral scanning mechanism for discrimination. Experimental validation on bighead carp demonstrates that adaptive progressive fusion and DS-DDPM augmentation improve classification accuracy by 2.1 and 3.5 percentage points over direct fourdomain combination and baseline training respectively. This advancement achieves 96.8% accuracy with only 5.83 M parameters for edge deployment, facilitating temporal-state dual-driven strategies for smart aquaculture transportation management. Meanwhile, it establishes a replicable paradigm for GAI applications addressing insufficient representation and data constraints in agricultural monitoring.

Heap fermentation is a characteristic stage of sauce-flavor Baijiu production and represents a complex open solid-state fermentation system shaped by gradients in temperature, oxygen availability, and moisture. Rather than a universal model, it is treated here as a well-documented case through which observations from an industrial fermentation can be organized and evaluated. This review synthesizes current knowledge on microbial succession, functional groups, and metabolite formation during heap fermentation and discusses these observations in relation to ecological assembly perspectives. The evidential scope and limitations of commonly used approaches are examined, including amplicon sequencing, quantitative assays, multi-omics analyses, spatial measurements, and modeling methods such as network analysis and reaction-diffusion simulations. Particular attention is given to uncertainties in linking community composition with process behavior and to practical constraints of applying advanced analytical tools in open fermentation environments. Studies have reported that environmental gradients are associated with shifts in microbial populations. Open inoculation, however, contributes variability among batches. Across published reports, some studies report relatively stable fermentation outcomes, although the mechanisms remain unclear. Accordingly, heap fermentation is discussed as a case-based context for a broader question relevant to many solid-state fermentations: why similar operating conditions yield reproducible outcomes in some cases but variable results in others. The review organizes available evidence, highlights current limitations and identifies key questions that may require further investigation.

The use of overnight cooked rice is known to enhance the texture of fried rice. To provide a theoretical foundation for its industrial production, this study examined the effects of 12-hour storage at 4 °C on the moisture content, textural properties, morphological characteristics, amylopectin-to-amylose ratio, thermal properties, and crystalline structure of rice. Scanning electron microscopy (SEM) revealed a hard, rough, polyhedral morphology, leading to the formation of debris. X-ray diffraction (XRD) showed an increase in relative crystallinity, indicating enhanced structural order from starch recrystallization. Differential scanning calorimetry (DSC) indicated higher retrogradation enthalpy and a greater degree of retrogradation (DR). Fourier transform infrared spectroscopy (FTIR) demonstrated an increase in the ratio of peak intensities at 1047/1022 cm⁻¹ (R1047/1022 cm⁻¹), confirming starch retrogradation and increased crystallinity. Concurrently, moisture content decreased, while hardness and chewiness increased; adhesiveness and springiness decreased. These findings suggest that overnight storage induces short-term retrogradation in rice, enhancing its suitability for fried rice preparation by improving its texture and appearance.

In industrial tofu production, soy protein gelation dictates final product quality, yet it is still monitored largely by operator judgement, leading to appreciable batch-to-batch inconsistency. Although ultrasonic techniques have shown considerable promise in tracking protein coagulation, most applications remain limited to static evaluation of current gelation state, which limits proactive control strategies. To overcome this gap, we propose a framework that integrates ultrasonic transmission sensing with multi-step time series forecasting. In particular, a triple-attention Seq2Seq (TA-Seq2Seq) model is developed, in which waveform attention identifies gelation-informative spectral regions, recency-biased temporal attention up-weights the most recent network development, and cross-attention to adapt information retrieval to each prediction horizon. Furthermore, a dual-probe through-transmission configuration with solid-state coupling pads was innovatively designed to mitigate the destabilising effect of high gelation temperatures on ultrasonic signal acquisition. To leverage ultrasonic information, full attenuation waveforms were collected at one-minute intervals over a 40-min gelation. Under Leave-One-Group-Out cross-validation, the experimental results demonstrate that the proposed model is capable of multi-step gelation degree forecasting across horizons from +1 to +5 min. For short-term prediction at the late gelation stage (t = 35 min), the model achieved a mean absolute percentage error as low as 1.02%, while for early-stage forecasting (t = 15 min), it still maintained an overall accuracy above 91.80%. Overall, by coupling ultrasonic signal trajectories with sequence modelling, this work advances soy protein gelation monitoring from static state estimation to multi-step forecasting, providing a quantitative basis for anticipatory endpoint determination and model-assisted process decision-making in tofu manufacturing.

Conventional food preservation materials often exhibit less/no biodegradability and multifunctionality, necessitating sustainable alternatives that combine active packaging with environmental safety. This study developed a multifunctional (biocompatible, nontoxic, high water-absorbable, sustain releasable, antioxidative/antimicrobial, and biodegradable) porous aerogel pad for food preservation (e.g. salmon). The aerogel pad was developed using sodium alginate (SA) and polyvinyl alcohol (PVA) as the structural matrix as well as green tea extract (GTE) as the active functional component. Systematic evaluation of GTE loading ratios (0-4.2 %) revealed that 3.0 % GTE exhibited a homogeneous three-dimensional interpenetrating network with the optimizing water absorption (2452.31 %), mechanical strength (hardness of 6308.61 g), and thermal stability. FTIR and SEM analyses confirmed a high-curvature pore structure formed between the GTE and matrix through hydrogen bonding interactions, which prolonged the antioxidant diffusion pathway as well as slowdown the release of antioxidant in aqueous- and lipid-rich simulants. In salmon preservation trials, the aerogel pads extended shelf life from 8 to 12 days, while soil degradation tests demonstrated complete biodegradation within 21 days. Above results together highlight that Alginate/PVA aerogel pad has potential as a sustainable, highperformance alternative for food packaging.

pH-sensitive controlled-release antimicrobial film was developed based on a zein-chitosan-Eudragit L100 matrix encapsulating vanillin (designated as ZCLV), aiming to provide sustained antimicrobial protection over the food storage phase, especially during the initial acidic spoilage period. The physical properties, microstructure, and proton-motive release behavior of film were thoroughly characterized. The ZCLV film exhibited superior moisture barrier properties, with a water vapor permeability (WVP) of 3.64 × 10-5 (g·m-1·s-1·Pa-1). It also demonstrated a high encapsulation efficiency (EE) of 53.82% and a loading content (LC) of 9.26%. Under test conditions at 4 °C, the ZCLV film provided an initial microbial reduction of 0.44 log cfu/g and sustained antibacterial activity, extending the shelf life of beef by 3 days. The dual pH-responsive release mechanism of the ZCLV film enables intelligent sequential release of vanillin that dynamically adapts to pH changes during beef spoilage, thereby achieving prolonged preservation.

Zhenjiang aromatic vinegar, one of the most renowned traditional Chinese cereal vinegars, is favored by consumers for its elegant and complex aroma. Aging is a critical process that significantly shapes its final flavor and quality. However, the succession of microbial communities and their impact on volatile flavor compounds during aging remain unclear. This study aimed to systematically investigate the dynamic changes and correlation analysis of volatile flavor compounds and microbial communities of Zhenjiang aromatic vinegar during aging. A total of 808 volatile flavor compounds were identified, and their contents differed after aging. The total content of volatile flavor compounds peaked at Y3, while the content of pyrans was highest at Y10. Twelve compounds contributed significantly to the overall aroma of Zhenjiang aromatic vinegar. With extended aging, the OAV of compounds like 2,4,5-trimethyloxazole increased, whereas those of others, such as α,α,4-trimethylcyclohex-3-ene-1-methylmercaptan decreased. Aging enhanced the unique aromatic properties and improved the flavor of Zhenjiang aromatic vinegar. High-throughput sequencing analysis revealed that Proteobacteria and Actinobacteriota were the dominant bacterial phyla during aging, with Firmicutes abundance peaking at Y3. Key bacterial genera in the fermentation stage, such as Pseudomonas, remained abundant during aging. Parengyodontium was the predominant fungal genus, while Meyerozyma exhibited phased changes during aging. Aspergillus and Penicillium emerged after a decrease in total acidity at Y5. Correlation analysis indicated that Aspergillus, Penicillium and Pseudomonas were significantly associated with approximately 75% of the key aroma compounds. Fungi might exert a greater influence on key aroma compounds than bacteria during aging. The findings deepen the understanding of flavor compounds and microbial communities in aged Zhenjiang aromatic vinegar, providing a theoretical basis for the optimization and quality control of aging.

This study investigated the effects of mechanical treatments, namely magnetic stirring (S), high-shear homogenization (HS), and ultrasonication (150, 300, and 600 W), on the sugarcane wax-based oil-in-water (O/W) emulsion and its integration into carboxymethyl cellulose (CMC)-based packaging films. The objective was to tackle issues, such as hydrophilic properties in CMC films and wax particle aggregation, by enhancing emulsion characteristics. Of the formulations, S/HS/300 W exhibited superior performance, attaining homogenous particle sizes ≤1 μm, a low polydispersity index (21.12), and a high ζ-potential (-41.55 mV), signifying improved stability. It also showed enhanced thermal stability, reduced crystallinity, and valuable rheology. Incorporating the S/HS/300 W emulsion into CMC films markedly improved water resistance, elevated hydrophobicity (115.5° water contact angle), decreased water vapor permeability by 23%, offered UV light protection dropped by 0%, decreased visible light transmission, and boosted mechanical qualities; the tensile strength rose by 34.2%. This study underscores the efficacy of mechanical processing in the production of high-performance films for sustainable packaging.

In meat quality monitoring, conventional destructive testing cannot meet modern supply chain demands for rapid, non-invasive assessment. To bridge this gap, Hyperspectral Imaging (HSI) stands out for combining spatial and spectral information. However, HSI alone exhibits inherent limitations, namely spectral overlap, surfaceonly detection, and limited access to volatile compounds. Multi-sensor data fusion based on HSI addresses these constraints. This review systematically examines HSI-based fusion through fusion architecture selection, Deep Learning (DL) integration, and dimensionality-driven fusion organization. The analysis first examines four architectures including low-level, mid-level, high-level, and hybrid data fusion. Subsequently, three intrinsic challenges of fusion, including modality heterogeneity, stage-wise information loss, and cross-modal contribution imbalance are identified. DL is positioned as the principal toolkit resolving these challenges. The review then examines branch-based fusion, which combines modality-specific deep features in a learned feature space, and model interpretability for auditable predictions. Furthermore, it introduces a dimensionality-driven framework that categorizes complementary modalities by data structure (1D, 2D, 3D), including electronic nose/Low Field Nuclear Magnetic Resonance (LF-NMR) (1D) for temporal biochemical dynamics, computer vision/multispectral/thermal imaging (2D) for spatial heterogeneity analysis, and dual-band HSI/Light Detection and Ranging (LiDAR)/Magnetic Resonance Imaging (MRI) (3D) for internal-external characterization. This framework consolidates existing applications, derives operational guidelines for fusion strategy selection, and reveals avenues for incorporating emerging sensing technologies. Future pathways including hybrid edge-cloud computing for real-time processing, cross-dimensional transfer learning for scalable modality integration, and miniaturized sensor arrays for cost-effective deployment are discussed. These directions position HSI-based fusion as a key technology for scalable meat quality monitoring.

Given that enantiomers can exhibit distinct biological activities, accurately measuring their content is of significant research importance. d-fructose is a sweetener widely used in food manufacturing in many countries, whereas its enantiomer l-fructose is biologically inert, affects normal human metabolism, and also poses challenges to food authenticity verification. In this study, plasmonic chiral nanoprobes were developed for surface-enhanced Raman scattering (SERS)-based enantioselective sensing of fructose. Herein, cysteine served as an inducer to synthesize gold nanoparticles with intricate surface structures. 4-Mercaptophenylboronic acid (4-MPBA) was introduced as a probe molecule. Leveraging the high binding affinity between 4-MPBA and fructose, fructose was effectively drawn into the near-field hotspot regions of the nanoparticles. This unique approach enabled the identification of fructose enantiomers even in complex environments. Notably, the complex surface architecture and needle-like tips of the nanoparticles contributed to exceptional electromagnetic enhancement performance, resulting in ultra-sensitive assay capabilities for fructose enantiomers. This study thus establishes an induction strategy for the rapid screening of food authenticity.

To enhance the preservation efficiency and pH-responsive functionality of active packaging, this study architecturally integrates antibacterial carbon dots (CDs) with tannic acid-Fe3+ metal-phenolic networks (MPNs) via coordination and hydrogen-bonding, and incorporates the assembled complex into a pectin (Pec) matrix to fabricate a pH-responsive composite film. The effects of CDs@MPNs incorporation on the microstructure, physicochemical properties, antibacterial performance, and biosafety of the resulting Pec/CDs@MPNs composite film were systematically investigated. The results demonstrated that CDs@MPNs functioned as multifunctional cross-linkers, improving the pectin network into a more compact and ordered architecture, which directly enhanced the film's water vapor barrier, mechanical strength, UV-blocking capacity, pH-responsive behavior, and biocompatibility. Furthermore, the composite film exhibited effective inhibitory activity against both Escherichia coli and Staphylococcus aureus. When applied to deep-fried meatballs during 13 d storage, the film achieved notable preservation effects: it reduced the peroxide value by 33.81%, the 2-thiobarbituric acid reactive substances value by 29.82%, and the total volatile basic nitrogen content by 28.73%. Biosafety assessments confirmed over 90% viability of mouse fibroblast cells and no observable abnormalities in zebrafish. This study successfully fabricates an integrated, biosafe, and naturally derived smart packaging material that offers an efficient strategy for food preservation.

Humidity serves as the most direct stimulus reflecting water balance and respiratory metabolism in fruit and vegetable cells. It offers higher real-time responsiveness and reversible control than temperature and light yet remain largely overlooked in current preservation technologies, limiting precise maintenance of produce freshness. To address this, this study constructs a humidity-responsive dual-surface Janus preservation pad. CQDs covalently graft onto Cu-MOF to yield CQDs@Cu MOF, which exhibits exceptional antioxidant activity (84.7 % DPPH radical scavenging) and broad-spectrum antibacterial efficacy. These composite particles integrate in situ into an electrospinning bilayer pad of EVOH and CA to form a hydrophilicity gradient. This design enables unidirectional water transport that prevents condensation and triggers smart release of active components under elevated humidity. Molecular dynamics simulations and docking analyses elucidate the conformational evolution of ethylene vinyl alcohol chains at different humidities and their role in regulating release kinetics. Storage trials show that lettuce using this Janus pad exhibits markedly reduced browning and sustains visual and nutritional quality for 16 days. This multiscale design strategy from molecular bonding and hierarchical structure assembly to functional performance validation provides an innovative and effective solution for intelligent and sustainable preservation of micro processed agricultural products.

Tofu is a traditional soy-based food of global importance, requiring stringent quality control to ensure safety and support industrial modernization. However, real-time monitoring in industrial-scale production is challenging because traditional methods are subjective, time-consuming, and destructive. In response to this challenge, a suite of rapid detection technologies based on optical, electrochemical, sensor, and physical field principles has emerged as a transformative solution. These technologies enhance detection speed, accuracy, and functionality, thus enabling the intelligent upgrading of the tofu industry. This review comprehensively summarizes and critically evaluates recent advancements in rapid detection methods for tofu, encompassing near-infrared spectroscopy, Raman spectroscopy, hyperspectral imaging, electronic nose/tongue, low-field nuclear magnetic resonance, ultrasonic testing, and photoelectrochemical sensors. For each technology, the underlying principles, application performance, and representative case studies are elucidated. In addition, persistent challenges such as model robustness, signal interference, and lack of standardization are analyzed. Finally, future trends are foreseen, with emphasis placed on multi-technology integration, intelligent automation driven by artificial intelligence, and the development of portable and online systems. This work aims to serve as a valuable reference for guiding future research and promoting the practical application of rapid detection technologies in the tofu industry.

To address the high diversity of Chinese dish categories, limited inter-class separability, and the scarcity of labeled samples in real-world scenarios, we propose a few-shot classification framework named the Multi-scale Prototype Neighbor Network (MPN-Net). Our approach builds upon Prototypical Networks, employing ResNet-18 as the backbone architecture and incorporating Efficient Channel Attention (ECA) together with a shared Transformer encoder to enhance channel-wise feature representation and global contextual modeling. Furthermore, we introduce a Prototype Neighbor Network (PNN) to integrate class prototypes with nearest-neighbor relationships between query and support instances, thereby enabling a dual-branch inference mechanism. Experimental evaluations on the Mini-ImageNet, UECFood-100, and VireoFood172 datasets demonstrate that MPN-Net achieves accuracies of 74.59%, 62.91%, and 74.05% under the 5-way 1-shot setting, and 86.06%, 87.32%, and 92.87% under the 5-way 5-shot setting, respectively, substantially outperforming baseline methods. The effectiveness of each component is further validated through comprehensive ablation studies and visualization analyses. Overall, our proposed framework provides a robust and effective solution for few-shot image classification in complex Chinese culinary scenarios.

Chinese yam (Dioscorea spp.) is widely used as food and traditional medicine, but varietal differences may affect dried yam slice quality. This study compared ten Chinese yam varieties, examined relationships between raw-material quality indicators and dried yam slice quality, and evaluated hyperspectral imaging for rapid assessment before processing. Significant varietal differences were observed in reducing sugar content, total phenolic content, and texture properties (p < 0.05), and trait–quality relationships were variety-dependent. Spectral preprocessing, variable selection, and regression modelling were used to predict the reference values of reducing sugar content and total phenolic content obtained using the specified analytical procedures, together with fresh-slice hardness. The best models combined principal component analysis with decision tree regression (PCA-DTR), competitive adaptive reweighted sampling with partial least squares regression (CARS-PLSR), and CARS with random forest regression (CARS-RFR), respectively. Prediction-set coefficients of determination (RP2) were 0.9891, 0.9335, and 0.9314, with root mean square errors of prediction (RMSEP) of 0.0981%, 0.0905 mg gallic acid equivalents/100 g dry weight, and 130.3973 gf, and residual predictive deviation (RPD) values of 9.7535, 3.9438, and 3.8820, respectively. These results support hyperspectral imaging with chemometrics for rapid assessment and selection of yam raw materials for dried yam slice processing.

Despite substantial progress in biosensor development, achieving reliable sensitivity and selectivity under real-world conditions remains challenging, particularly in complex and heterogeneous sample matrices. While sensitivity has historically been the primary focus of biosensor optimization, selectivity often emerges as a limiting factor for practical performance when deployed outside controlled laboratory environments. Poor selectivity can lead to false-positive or false-negative results, thereby undermining the reliability and accuracy of biosensing platforms. A major contributor to this problem is nonspecific binding, the unintended interaction between biosensor and nontarget species. However, the origins, mechanism, and implications of nonspecific binding remain insufficiently understood and are still actively debated within the scientific community. In this review, we trace the conceptual development of nonspecific binding and critically examine its physicochemical origins in substrates and biorecognition elements. We then assess recent progress in recognition elements, such as antibodies, aptamers, and enzymes, emphasizing not only their strengths but also their limitations and vulnerability to off-target interactions. To mitigate nonspecific binding, we summarize a range of emerging strategies, including optimizing the conjugation and orientation and increasing binding site accessibility and density through structural design, removing interfering species, and implementing signal-level strategies. Finally, we outline persisting challenges and future directions for enhancing biosensor selectivity. Collectively, these insights offer a roadmap for designing next-generation biosensors with high accuracy, robust selectivity, and real-world applicability.

Ensuring food safety and quality has emerged as a critical global priority, driven by rapid growth of the world population and the increasing complexity of modern food production and distribution systems. As supply chains become more globalized and diversified, risks associated with contamination, adulteration, and quality degradation have intensified, posing significant challenges to public health and international trade. Conventional detection and monitoring methods, although scientifically established, often face limitations in efficiency, cost-effectiveness, and adaptability to diverse contamination sources and dynamic production environments. In recent years, the emergence of deep learning as a transformative branch of artificial intelligence has opened new opportunities to address these limitations through automated data analysis and intelligent pattern recognition. This review provides a comprehensive synthesis of advances in artificial intelligence applications for food safety and quality control from 2019 to 2025, covering hazard detection, quality evaluation, and intelligent monitoring for real-time risk prediction and decision support. By systematically analyzing representative methodologies, technical frameworks, and performance outcomes, the review underscores the advantages of artificial intelligence in achieving high-throughput, non-destructive, and precise analytical performance across complex food matrices. It also identifies persistent challenges, including data imbalance, limited interpretability, and environmental variability, which hinder large-scale deployment and regulatory integration. The convergence of artificial intelligence with advanced sensing technologies, big data analytics, and domain expertise is expected to drive the evolution of next-generation intelligent food safety systems, ultimately enabling more transparent, adaptive, and sustainable approaches to postharvest processing, quality management, and global food security.