Transformative artificial intelligence integration in aquatic supply chains: synergizing precision aquaculture with intelligent logistics and data-driven consumption.

Transformative artificial intelligence integration in aquatic supply chains: synergizing precision aquaculture with intelligent logistics and data-driven consumption - PMC Skip to main content An official website of the United States government Here's how you know Here's how you know Official websites use .gov A .gov website belongs to an official government organization in the United States. Secure .gov websites use HTTPS A lock ( Lock Locked padlock icon ) or https:// means you've safely connected to the .gov website. Share sensitive information only on official, secure websites. Search Log in Dashboard Publications Account settings Log out Search… Search NCBI Primary site navigation Search Logged in as: Dashboard Publications Account settings Log in Search PMC Full-Text Archive Search in PMC Journal List User Guide PERMALINK Copy As a library, NLM provides access to scientific literature. Inclusion in an NLM database does not imply endorsement of, or agreement with, the contents by NLM or the National Institutes of Health. Learn more: PMC Disclaimer | PMC Copyright Notice Food Chem X . 2026 Apr 5;35:103833. doi: 10.1016/j.fochx.2026.103833 Search in PMC Search in PubMed View in NLM Catalog Add to search Transformative artificial intelligence integration in aquatic supply chains: synergizing precision aquaculture with intelligent logistics and data-driven consumption Xiaonan Fan Xiaonan Fan a School of Management, Dalian Polytechnic University, Dalian 116034, Liaoning, China. Find articles by Xiaonan Fan a , Jiyu Zou Jiyu Zou a School of Management, Dalian Polytechnic University, Dalian 116034, Liaoning, China. Find articles by Jiyu Zou a , Yang Liu Yang Liu b SKL of Marine Food Processing & Safety Control, National Engineering Research Center of Seafood, Collaborative Innovation Center of Seafood Deep Processing, Liaoning Province Key Laboratory for Marine Food Science and Technology, School of Food Science and Technology, Dalian Polytechnic University, Dalian 116034, China. Find articles by Yang Liu b, ⁎ , Dongmei Li Dongmei Li b SKL of Marine Food Processing & Safety Control, National Engineering Research Center of Seafood, Collaborative Innovation Center of Seafood Deep Processing, Liaoning Province Key Laboratory for Marine Food Science and Technology, School of Food Science and Technology, Dalian Polytechnic University, Dalian 116034, China. Find articles by Dongmei Li b , Dayong Zhou Dayong Zhou b SKL of Marine Food Processing & Safety Control, National Engineering Research Center of Seafood, Collaborative Innovation Center of Seafood Deep Processing, Liaoning Province Key Laboratory for Marine Food Science and Technology, School of Food Science and Technology, Dalian Polytechnic University, Dalian 116034, China. Find articles by Dayong Zhou b , Hai Chi Hai Chi c Key Laboratory of Protection and Utilization of Aquatic Germplasm Resource, Liaoning Ocean and Fisheries Science Research Institute, Dalian 116023, China. Find articles by Hai Chi c , Deyang Li Deyang Li b SKL of Marine Food Processing & Safety Control, National Engineering Research Center of Seafood, Collaborative Innovation Center of Seafood Deep Processing, Liaoning Province Key Laboratory for Marine Food Science and Technology, School of Food Science and Technology, Dalian Polytechnic University, Dalian 116034, China. Find articles by Deyang Li b, ⁎ Author information Article notes Copyright and License information a School of Management, Dalian Polytechnic University, Dalian 116034, Liaoning, China. b SKL of Marine Food Processing & Safety Control, National Engineering Research Center of Seafood, Collaborative Innovation Center of Seafood Deep Processing, Liaoning Province Key Laboratory for Marine Food Science and Technology, School of Food Science and Technology, Dalian Polytechnic University, Dalian 116034, China. c Key Laboratory of Protection and Utilization of Aquatic Germplasm Resource, Liaoning Ocean and Fisheries Science Research Institute, Dalian 116023, China. ⁎ Corresponding author: School of Food Science and Technology, Dalian Polytechnic University, Dalian 116034, China. [email protected] [email protected] Received 2025 Dec 9; Revised 2026 Mar 9; Accepted 2026 Apr 4; Collection date 2026 Apr. © 2026 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). PMC Copyright notice PMCID: PMC13092574  PMID: 42016730 Abstract Aquatic supply chains currently encounter critical sustainability and efficiency challenges, ranging from resource overexploitation to inherent cold chain vulnerabilities. Although individual artificial intelligence (AI) applications are emerging, research synthesizing the entire “Farm-to-Table” continuum remains scarce. This review bridges this gap by evaluating AI integration—specifically machine learning and deep learning—across aquaculture, harvesting, processing, logistics, and marketing. The analysis reveals that while AI demonstrates notable efficacy in precision tasks like dynamic water quality prediction and automated catch classification, applications in pre-processing and low-altitude delivery remain nascent. Future advancements require interpretable algorithms, standardized databases, interdisciplinary collaboration, and cost-effective deployment to construct resilient, intelligent supply chains that ensure food safety and satisfy growing global market demands. Ultimately, this review provides a robust theoretical foundation for researchers and practitioners to enhance product safety and operational efficiency, fostering a sustainable, digital transformation of the aquatic industry. Keywords: Artificial intelligence, Aquatic products, Supply chain, Machine learning, Deep learning, Quality and safety Highlights • AI optimizes end-to-end aquatic chains via precision aquaculture & smart logistics. • DL achieves high accuracy in water quality forecasting and disease diagnosis. • Vision-AI behavioral analytics reduces feeding waste in aquaculture. • SERS-1D CNN enables reliable detection of antibiotic residues for safety inspection. • Blockchain-AI synergies enhance cold-chain transparency and minimize quality loss. 1. Introduction Aquatic products constitute a vital component of human dietary health, characterized by high protein and essential micronutrient content. However, they concurrently harbor foodborne hazards including pathogenic bacteria, viruses, parasites, heavy metals, and environmental contaminants ( Shen et al., 2025 ). Ensuring their quality and safety has thus emerged as a critical concern. The aquatic product supply chain encompasses culture, capture, processing, transportation, storage, and retail, with any disruption potentially compromising product integrity ( Wei et al., 2024 ). Compared to those of other food categories, the journey from harvest to consumption involves greater complexity and stricter requirements for environmental control, technical precision, and cost management. Consequently, achieving stable, efficient, and transparent supply chains has garnered significant academic and industrial attention ( Cromwell et al., 2025 ; Jiao et al., 2025 ). Current approaches employing automated equipment, digital management systems, and interconnected decentralized optimization have partially addressed sectoral challenges ( De & Roy, 2024 ). Nevertheless, limitations persist in data integration depth, dynamic decision-making capabilities, personalized service accuracy, and end-to-end coordination efficiency. Artificial intelligence (AI), encompassing technologies defined as self-learning adaptive systems, can be classified by application, functionality, or agent type. Its evolution spans machine learning (ML), natural language processing, computer vision, and deep learning (DL), progressing toward embodied intelligence. AI applications now extend across agriculture, healthcare, finance, transportation, manufacturing, and education ( Javaid et al., 2023 ). Within crop supply chains specifically, AI enables key operations: during cultivation, it facilitates soil composition analysis ( Kaplan et al., 2024 ), climate forecasting ( Zhao et al., 2024 ), planting strategy optimization ( Linaza et al., 2021 ), pest/disease prediction ( Kariyanna & Sowjanya, 2024 ), irrigation management ( Pallathadka et al., 2022 ), and growth monitoring ( Kollia et al., 2021 ). Post-harvest, AI supports crop identification ( Kang et al., 2020 ), freshness detection ( Kang et al., 2022 ), and quality assessment ( Ali et al., 2021 ). In storage and transportation, it enhances inventory management ( Misra et al., 2022 ) and route optimization ( Zhou et al., 2023 ). AI further enables nutritional evaluation and quality certification in sales ( Misra et al., 2022 ). These applications demonstrate AI’s predictive, control, and reasoning capabilities in crop supply chains. Given structural and operational parallels between crop and aquatic product supply chains, this paradigm offers transferable insights. Additionally, AI applications in food science encompass produce safety inspection ( Yang et al., 2025 ), foodborne pathogen detection ( Deng et al., 2025 ), and shelf-life prediction. To date, a series of scholarly overviews have emerged focusing on the integration of AI within the realm of food safety. These studies predominantly categorize AI’s contributions into pivotal areas such as the identification of potential risk factors ( Yumnam et al., 2024 ), high-precision safety screening ( Yuan et al., 2025 ), and the detection of food fraud and adulteration ( Wang, Gu, et al., 2024 ). While related reviews proliferate, there is a conspicuous absence of relevant reviews addressing AI-enabled monitoring of safety and quality dynamics throughout the entire aquatic products continuum. Current scholarly discussions often lack a cohesive perspective that spans initial production and industrial processing to the end-consumer phase. Furthermore, existing literature frequently adopts a generalized approach toward food categories, failing to account for the unique requirements of diverse aquatic species. This oversight is particularly significant given that the efficacy of AI models is profoundly influenced by the distinct biological properties of raw materials and the intricate, high-stakes logistics characteristic of aquatic supply chains. Therefore, this review provides a comprehensive synthesis of AI-driven advancements in safeguarding the quality and safety of aquatic products across the entire “Farm-to-Table” continuum, encompassing aquaculture and harvesting, processing, logistics, and marketing. Taken together, AI-enabled aquatic product supply chains represent an interdisciplinary nexus of computer science, food science, and management science. This integration addresses sector-specific and systemic challenges to enhance efficiency and ensure quality throughout the supply chain. From a food science perspective, this review synthesizes technological breakthroughs in AI-driven quality control, including precise contaminant identification via spectroscopic analysis coupled with DL, dynamic freshness assessment through multi-modal data integration, and predictive models for quality deterioration during processing ( Fig. 1 ). These advances provide cross-disciplinary solutions to core challenges of perishability and safety, catalyzing a paradigm shift from experience-based to data-driven food inspection. Management science insights reveal AI’s restructuring effects on supply chain operations ( Table 1 ). Demand forecasting algorithms optimize inventory turnover, intelligent cold-chain scheduling reduces loss, and traceability systems strengthen supply-demand trust. Collectively, these innovations establish actionable frameworks for operational synergy, cost efficiency, and sustainable development. Furthermore, this review bridges theoretical research and industrial implementation. It delineates emerging frontiers in bio-feature engineering and food chemistry for researchers, while furnishing practitioners with viable pathways for technology adoption. This work provides critical guidance for advancing intelligent and digital transformation within the aquatic products industry. Fig. 1. Open in a new tab Integrated AI applications across the aquatic product supply chain. Table 1. AI models applicable to various stages of the aquatic product supply chain and their application maturity Relevant stages of aquatic products supply chain AI models AI application maturity References Aquaculture Water quality monitoring ANN, CNN, SVR-RGA, LSTM, IPSO-LSSVR Local application is mature, and the whole is still in the optimization stage ( Guo, Dong and Lee, 2020 , Yang et al., 2022 ) Disease diagnosis CNN, RFC Single species diagnosis of fish is mature, and multi- species complex environment needs to be expanded ( Waleed et al., 2019 ; González et al., 2023 ) Feeding decision-making MEA-BPNN, ANFIS, CNN, DSRN, Faster MSSDLite Precision feeding technology is close to industrialization ( Chen et al., 2019 ; Zhou et al., 2019 ) Capture Underwater harvesting ResNet, LinkNet-34, GAN, MobileNet-V2, SSD, YOLOv5, CNN Single species, mature static scene, weak complex environment ( Wang, Guo, et al., 2022 ; Zhang, Wang, Zhang, Li and Sun, 2023 , Zhang, Yue, Song, Jia and Li, 2023 ) Catch classification DAGM-SVM, VBM-SVM, NN, BPNN, AdaBoost, CNN, YOLOv5, Faster R-CNN Lack of accurate classification in multi-species mixed scenes ( Hu et al., 2012 ; Zhang, Wang, et al., 2023 ) Processing Pollution detection 1-D CNN, CNN, AIHazardsFinder Spectrum+AI paradigm standardized ( Maurya et al., 2023 ; Teng et al., 2023 ) Freshness assessment ANN, k-NN, CNN, SR Lack of generic models across species ( Atasoy et al., 2015 ; Wu et al., 2022 ) Pre-processing None Lack of intelligence in the whole process None Cutting DNN, CNN, k-means Lack of multi-species adaptability ( Mainali, 2021 ; Wang et al., 2018 ) Deep processing ANN Lack of standardized process database ( Hosseinpour et al., 2014 ; Sadhu et al., 2020 ) Cold chain logistics Demand forecasting Bi-LSTM Less AI models applied ( Aldahmani et al., 2024 ; Huber et al., 2017 ) Optimization of transport equipment and route planning RNAeACO Lack of data sets and few specific application cases of AI ( Fallmann et al., 2023 ; Zhang, Tseng, et al., 2019 ) Temperature control ANN, LSTM, DNN Remarkable results have been achieved ( Loisel et al., 2022 ; Zou et al., 2023 ) Cold storage control LSTM, k-means, CNN Multi system collaborative optimization has not yet been realized ( Eze et al., 2024 ; Liu et al., 2022 ) Frozen-thawed monitoring None AI model of anhydrous product adaptation None Energy consumption optimization ANN Less AI models applied ( Ribault et al., 2021 ; Zhu et al., 2023 ) Low-altitude delivery None The technology is feasible but needs further optimization and promotion None Sales Authenticity detection CNN, DNN, SVM, LR, 1-D CNN There are many AI application cases ( Currò et al., 2024 ; Varrà et al., 2022 ) Sales Forecasting and Quality Control BP-LSTM-Verhulst, Faster R-CNN Sales forecasting model lacking multi-source data fusion ( Tian & Wang, 2022 ; Sykes, 2022 ) Personalized recommendation None Insufficient research accumulation is still in its infancy None Open in a new tab The abbreviations of AI models in Table 1 are expanded as follows: ANN: Artificial Neural Network; CNN: Convolutional Neural Network; SVR-RGA: Support vector regression optimized by real-coded genetic algorithm; LSTM: Long short-term memory; IPSO-LSSVR: Improved particle swarm optimization-least squares support vector regression; RFC: Random forest classifier; MEA-BPNN: Mind evolutionary algorithm-Back Propagation Neural Network; ANFIS: Adaptive neuro-fuzzy inference system; DSRN: Dual-stream recurrent network; Faster MSSDLite: Faster mobile single shot multiBox detector lite; ResNet: Residual network; LinkNet-34: “A lightweight semantic segmentation network”; GAN: Generative adversarial network; MobileNet-V2: MobileNet version 2; SSD: Single shot multiBox detector; YOLOv5: “You only look once version 5”; DAGM-SVM: Directed acyclic graph method-support vector machine; VBM-SVM: Voting based method support vector machine; NN: Neural network; BPNN: Back Propagation Neural Network; AdaBoost: Adaptive boosting; faster R-CNN: Faster region-based Convolutional Neural Network; 1-D CNN: “One-dimensional Convolutional Neural Network”; AIHazardsFinder: Artificial intelligence hazards finder; k-NN: k-nearest neighbors; SR: Stochastic resonance; DNN: Deep neural network; k-means: k-means clustering; Bi-LSTM: Bidirectional long short-term memory; RNAeACO: Ribonucleic acid-inspired ant colony optimization; SVM: Support vector machine; LR: Logistic regression; BP-LSTM-Verhulst: BP-LSTM combined with verhulst model. 2. Intelligent Optimization in Aquatic Farming and Harvesting As the foundational stage of aquatic product supply chains, farming and harvesting operations are increasingly enhanced by AI. This integration fully addresses inefficiencies and experience-dependency in established paradigms through data-driven precision modeling, computer vision-based intelligent analysis, and decision optimization via multi-modal integration. Substantial innovations have been achieved in five domains: water quality monitoring, disease diagnosis, feeding strategy optimization, underwater harvesting, and catch classification ( Fig. 2 A). Fig. 2. Open in a new tab AI applications in aquaculture and fisheries: (A) Comprehensive intelligent innovation across five aquaculture and capture domains; (B) Harmful algae identification framework ( Yang et al., 2022 ) Copyright 2022, Springer; (C) Automated fish disease recognition architecture ( Waleed et al., 2019 ) Copyright 2019, IEEE; (D) Dual-model feeding decision system ( Jang, 1993 Copyright 1993, IEEE; Zhou et al., 2018 Copyright 2018, Elsevier); (E) Deep learning-based live shrimp recognition ( Wang, Guo, et al., 2022 ) Copyright 2022, IEEE; (F) Ark shell classification models ( Kim et al., 2024 ). 2.1. AI Applications in Aquaculture 2.1.1. Water Quality Monitoring Aquatic ecosystems represent critically endangered environments globally, partially attributed to anthropogenic pollutant discharge from industrialization, urbanization, and aquaculture. Such discharges induce rapid nutrient enrichment (e.g., nitrogen, phosphorus), triggering algal blooms through accelerated proliferation. As hallmark manifestations of eutrophication, these blooms profoundly impact aquatic ecosystems: nocturnal respiration by algae and heterotrophic microorganisms rapidly depletes dissolved oxygen (DO), while microbial decomposition of senescent algal biomass further reduces DO below 2 mg/L, establishing hypoxic conditions. Certain blooms (e.g., cyanobacteria- or dinoflagellate-derived) involve toxigenic species. Exposure to elevated toxin concentrations induces acute piscine toxicity, manifesting as branchial lesions, hepatic hemorrhage, and mass mortality events. Traditional algal bloom mitigation primarily relies on physical intervention, chemical agents, and biological regulation. Li et al. (2022) employed full factorial design and response surface methodology to investigate cumulative effects of flow velocity, copper sulfate, and propionamide on Microcystis growth. Their findings demonstrated synergistic algal suppression with chemical-biological combinations, whereas antagonistic interactions emerged upon introducing physical measures, with flow velocity exhibiting dominant influence. This provides an empirical basis for optimized control strategies. Although these approaches demonstrate partial efficacy, their heavy reliance on manual monitoring substantially elevates operational costs. Concurrent limitations in dynamic data acquisition impede real-time tracking of bloom evolution, constraining practical applicability and effectiveness. AI technologies effectively mitigate inherent constraints of conventional methods. To establish a foundation for real-time monitoring, Guo et al. (2020) subsequently developed a real-time data-driven prediction system integrating vertical stability theory with Artificial Neural Network (ANN). Validated against 191 bloom events, it achieved daily risk forecasting with high accuracy. To further enhance the applicability of such systems across diverse geographical regions, Park et al. (2025) applied a transfer learning (TL) framework using Transformer-based architectures to enhance harmful algal bloom (HAB) forecasting across 26 different river sites. Their approach successfully addressed data scarcity and improved average prediction accuracy, demonstrating the robust generalizability of TL in diverse aquatic environments. Building upon these predictive frameworks, recent research has integrated heterogeneous data sources to refine severity assessments. Zhao et al. (2023) developed a multi-modal DL framework to evaluate the severity levels of HABs. Their approach outperformed state-of-the-art unimodal methods, achieving a RA-RMSF of 0.816 in identifying high-severity bloom events. Beyond broad severity levels, increasing the granularity of forecasting targets has become a critical focus for precise management. Ahn et al. (2023) developed a forecasting model by integrating Convolutional Neural Network (CNN) with transformer and temporal fusion transformer architectures to directly predict harmful cyanobacterial cell counts. Their approach shifted the focus from traditional chlorophyll-a indicators to species-specific density, achieving robust short-term predictions that facilitate more precise operational algae warning systems. Finally, to automate the identification process at the taxonomic level, Yang et al. (2022) employed CNNs and TL to automate harmful algal classification. Their method improved average accuracy by 11.9% in distinguishing 11 toxic and 31 non-toxic algal species, significantly reducing expert labor. This technological evolution reflects an advancing trajectory: from single-system forecasting to multi-scenario real-time warning, culminating in microscopic species identification ( Fig. 2 B). Collectively, these studies establish complementary capabilities across macroscopic trend analysis, mesoscale risk alerting, and micro-level taxonomic discrimination. Algal blooms represent a singular manifestation of water pollution within a multi-dimensional monitoring framework encompassing diverse contaminants. AI technologies significantly enhance real-time analysis and intelligent early-warning capabilities for multivariate data, improving detection efficiency for dynamic environmental variations and predictive capacity for anomalous risks. This provides critical technical support for precision regulation of aquatic ecosystems. Liu, Tai, et al. (2013) developed a hybrid methodology integrating support vector regression (SVR) with real-coded genetic algorithm optimization for aquaculture water quality forecasting. Empirical validation demonstrated superior accuracy and generalization in DO and temperature prediction compared to conventional SVR and backpropagation neural networks (BPNNs). Distinct AI models exhibit complementary strengths: long short-term memory (LSTM) networks achieve precise forecasting of critical parameters including DO, demonstrating robust predictive performance ( Zhao et al., 2021 ). The improved PSO-least squares support vector regression (IPSO-LSSVR) hybrid model outperformed standard SVR and BPNNs in DO prediction for crab aquaculture, providing an effective solution for intensive systems ( Liu, Xu, et al., 2013 ). Collectively, SVR-RGA and IPSO-LSSVR exemplify algorithmic refinements within traditional ML frameworks, enhancing model efficacy through optimization techniques suitable for medium/small datasets. Conversely, LSTM leverages DL advantages for superior feature extraction from large-scale temporal data, proving more effective for long-term dynamic prediction in complex environments. This progression signifies a technological transition from heuristic-optimized algorithms to automated DL modeling in water quality monitoring. 2.1.2. Disease Diagnosis Conventional diagnosis of aquatic diseases is time-consuming, experience-dependent, and spatially constrained. This approach exhibits low efficiency and fails to meet rapid response requirements in field-level aquaculture operations. AI-powered computer vision and DL enable: (1) Identification of pathognomonic features on organisms’ surfaces/tissues and abnormal behavioral patterns; (2) Detection of subclinical infections, microscopic pathogen characteristics, and disease-associated biomarkers in aquatic environments; (3) Quantitative assessment of individual health impairment, population infection rates, and transmission risks. Integration with environmental data further facilitates holistic health status evaluation and disease progression forecasting, providing a quantitative basis for early intervention and precision control. Current AI applications predominantly focus on piscine disease diagnosis. Waleed et al. (2019) developed a computer vision-CNN system automating detection of epizootic ulcerative syndrome, ichthyophthiriasis, and columnaris disease through behavioral image analysis, achieving 99.04% accuracy ( Fig. 2 C). González et al. (2023) implemented semi-supervised learning for matrix-assisted laser desorption/ionization time-of-flight mass spectrometry data to identify Piscirickettsia salmonis infections in salmonid aquaculture. Their self-training algorithm integrated with random forest classifier attained 90% accuracy and 75% sensitivity, surpassing conventional supervised approaches. AI algorithms further enable extraction of multi-dimensional biomarkers from fish health data. ML-driven pattern recognition precisely detects behavioral anomalies (e.g., swimming trajectories, schooling distributions) and feeding irregularities (e.g., frequency, feed conversion efficiency), facilitating early identification of subclinical pathologies or stress responses. These innovations overcome spatiotemporal constraints and efficiency limitations inherent in manual diagnosis, establishing novel paradigms for preemptive intervention. Nevertheless, current research remains concentrated on singular piscine species. Further exploration of technical robustness in complex aquaculture environments and multi-species/pathogen diagnostic capacities is essential to advance AI implementation in aquatic disease management. 2.1.3. Feeding Strategy Optimization Feeding strategies critically impact aquaculture profitability and sustainability. Conventional approaches relying on empirical judgments often cause feed waste, water pollution, and uneven growth. AI technologies overcome these limitations by integrating sensor networks, image recognition, big data analytics, and ML to monitor organismal growth and feeding behaviors in real-time, enabling precision feeding models. Chen et al. (2019) developed a mind evolutionary algorithm-backpropagation neural network (MEA-BPNN) model predicting feed intake in intensive fish farming with high accuracy, providing theoretical foundations for intelligent feeding systems. Zhou et al. (2018) created an infrared computer vision-based system coupled with adaptive neuro-fuzzy inference system (ANFIS), extracting fish aggregation indices and feeding intensity metrics to achieve 98% decision accuracy. This approach improved feed conversion by 10.77% and significantly reduced pollution versus scheduled feeding protocols ( Fig. 2 D). Further advancing this field, Zhou et al. (2019) implemented a CNN-machine vision method for automated feeding intensity classification (90% accuracy). For salmonid applications, Måløy et al. (2019) designed a dual-stream recurrent network (DSRN) integrating “two-dimensional Convolutional Neural Network” spatial feature extraction and “three-dimensional Convolutional Neural Network” motion analysis with LSTM sequence classification, attaining 80% accuracy in feeding behavior prediction—surpassing conventional CNN and CNN-LSTM hybrids. Addressing crustacean aquaculture, Cao et al. (2020) deployed Faster mobile-SSD-lite for real-time crab distribution detection in complex underwater environments. Achieving 99.01% mean precision at 74.07 frames per second, this model outperformed traditional methods in accuracy and speed. This evolutionary trajectory demonstrates AI’s paradigm shift from passive environmental response to active behavioral perception. Continuous advancements—from MEA-BPNN environmental integration → ANFIS static feature utilization → DSRN dynamic behavior capture → lightweight architectural efficiencies—progressively balance precision with real-time operational demands. 2.2. AI Applications in Harvesting Operations 2.2.1. Underwater Harvesting Traditional manual diving operations incur high costs, safety hazards, and inefficiencies. Intelligent mechanized harvesting represents the developmental trajectory. AI technologies drive the transition from experience-dependent practices to precision operations through enhanced capabilities in: (1) Accurate positioning; (2) Target recognition; (3) Dynamic decision-making. This technological progression extends from static feature extraction to dynamic target tracking and control, continuously adapting harvesting protocols across diverse aquatic species and scenarios. Sea cucumber harvesting poses significant challenges due to morphological and locomotive characteristics. Wei et al. (2016) developed a vision-based tracking algorithm employing defogging techniques to enhance target-background contrast and mean-shift optimization for precise localization, improving robustness in complex underwater environments. Qiao et al. (2017) implemented an active contour model for automated segmentation, achieving 96.54% accuracy across 120 samples with average processing time of 4.27 seconds, outperforming conventional methods for real-time harvesting operations. Advancing beyond dynamic tracking and static contouring, Zhang, Yu, et al. (2020) enhanced deep-sea robustness via stochastic gradient descent-optimized detection. Utilizing c-watch underwater robots for image acquisition coupled with deep residual networks, their method attained 98.8% recognition accuracy on 120 images, demonstrating superior practicality and precision over traditional approaches. AI technologies demonstrate significant value in intelligent fish harvesting. Hao et al. (2022) engineered an intelligent fishing system integrating programmable logic controller with fuzzy proportional-integral-derivative control, exhibiting enhanced stability and dynamic performance for scientifically optimized operations. Pourkabirian et al. (2023) developed a hybrid localization method combining angle of arrival and received signal strength for underwater internet of things (IoT)-enabled harvesting. Compared with benchmark approaches, this method improves positioning accuracy by >13%, enabling targeted fish capture while mitigating bycatch of endangered species, reducing energy consumption, and lowering fuel costs. Complementing these innovations, Konovalov et al. (2019) established an image-based weight estimation framework. Their LinkNet-34 CNN automates fish image segmentation, with subsequent area-to-weight modeling achieving 4.36% mean absolute percentage error using fin-excluded univariate linear regression, demonstrating robust predictive capability. Current AI applications in shrimp and shellfish harvesting remain limited, yet precise species identification is critical for targeted capture. Wang, Guo, et al. (2022) developed a real-time recognition framework using modified generative adversarial network for image enhancement with a lightweight MobileNet-V2 single shot multibox detector model, achieving 90.32% accuracy. This enables real-time monitoring of shrimp distribution and density for informed harvesting decisions ( Fig. 2 E). Complementing this, Hu, Chen, et al. (2023) implemented “you only look once version 5 (YOLOv5)” for Litopenaeus vannamei body length and feed consumption assessment. Image enhancement techniques improved detection in turbid environments, while real-time size monitoring optimized harvest timing to maximize economic returns at ideal specifications. For shellfish, Zhang, Yue, et al. (2023) designed feature library network—a CNN-based identification system for morphologically similar species. Utilizing a 68-species image dataset with filter pruning/repair modules and hybrid loss functions, it attained 93.95% accuracy, enabling species-selective harvesting to mitigate bycatch and ecological impacts. These studies—spanning visual tracking and automated cutting for sea cucumbers, optimized deep-sea detection models, precision control in fish harvesting, and species-specific identification for shrimp/shellfish—demonstrate AI’s evolutionary trajectory in aquatic harvesting. Applications are transitioning from static feature recognition toward dynamic behavioral perception and complex environmental adaptation. Methodologically, advancements progress from traditional algorithm-based image enhancement and segmentation to optimized DL architectures, culminating in lightweight model deployment and heterogeneous cluster coordination. This progression addresses technical constraints while enhancing operational efficiency and precision, progressively fulfilling real-time processing and robustness requirements in challenging underwater environments. 2.2.2. Catch Classification Traditional catch classification systems rely heavily on empirical judgment, resulting in low efficiency and inconsistent outcomes. AI technologies—particularly ML and DL—enable automated species identification post-harvest. For piscine classification, Hu et al. (2012) developed a feature-based identification method using color/texture attributes with multi-class support vector machine (SVM). Their directed acyclic graph multi-class support vector machine achieved 97.77% mean accuracy at 12.7 ms processing time, demonstrating superior efficiency. Complementing this, Cao et al. (2021) engineered a computer vision-neural network system classifying nine fish species at 93% accuracy. Zhang, Li, et al. (2020) advanced automated counting via image density grading and local regression, constructing BPNN-based models for precise enumeration. This approach outperformed Cao’s method in stability and real-time processing. For other economically significant aquatic species, Zhang et al. (2014) implemented an evolutionary constructed features method. This approach automatically generates features via genetic algorithms and classifies shrimp integrity using AdaBoost, achieving 95.1% accuracy with robust adaptability. Hu, Wu, et al. (2020) developed ShrimpNet, a CNN-based architecture attaining 95.48% identification accuracy, demonstrating operational efficacy. Later, Zhang, Wang, et al. (2023) engineered a YOLOv5-(2D) 2 PCA framework for Chinese mitten crab carapace recognition. Initial detection reached 99.9% precision via TL, followed by 84.42% identification accuracy at 1.859 seconds matching time. Complementing these, Hu, Zhou, et al. (2020) created a rapid squid classification and freshness assessment system using modified Faster Recurrent CNNs, attaining 85.7% mean accuracy in 0.144 seconds. Feng et al. (2021) enhanced Faster region-based Convolutional Neural Network (R-CNN) for shellfish recognition, improving multi-target detection accuracy by nearly 4% in complex backgrounds. Kim et al. (2024) established a CNN-based model classifying three ark shell phenotypes with 92.4% accuracy, providing theoretical foundations for bivalve taxonomy ( Fig. 2 F). Collectively, AI technologies have transformed classification practices from experience-dependent operations to automated intelligent systems. This evolution spans from feature engineering refinement in fish sorting to architectural innovations for multi-species classification (shrimp, shellfish, squid). Methodologically, the progression from conventional feature extraction to CNN implementation, and further to lightweight real-time architectures, progressively meets the dual demands for precision and rapid response in complex sorting scenarios. These advancements significantly enhance operational efficiency and sorting accuracy across aquatic product categories ( Table 2 ). Table 2. Application of AI model in catch classification Model Application AI-inspired creativity Learning Type Model performance References Machine Learning MSVM Grass carp, blackfish, Wuchang fish and pomfret Accurate species identification of fish with mobile devices Supervised learning Real-time response model, Lightweight model, Low concurrency model ( Hu et al., 2012 ) BPNN Flatfish Accurate identification and automatic recording and tracing of fish species and length data Supervised learning Delayed response model, Lightweight model, Low concurrency model ( Cao et al., 2021 ) Fish Automatic fish counting based on image density classification and local regression Supervised learning Real-time response model, Lightweight model, Low concurrency model ( Zhang, Li, et al., 2020 ) AdaBoost Shrimp Efficient and high accuracy adaptive machine vision classification Supervised learning Real-time response model, Lightweight model, ( Zhang et al., 2014 ) Deep Learning ShrimpNet White shrimp Efficient and accurate automatic recognition using deep learning Supervised learning Delayed response model, Lightweight model, Low concurrency model ( Hu et al., 2020 ) improved Faster R-CNN Todarodes pacificus Fast, low-cost, non-intrusive automated classification and freshness assessment integration Supervised learning Real-time response model, Lightweight model, Low concurrency model ( Hu, Zhou, et al., 2020 ) Faster R-CNN Clam, Mussel Achieve accurate recognition and positioning of multiple targets Supervised learning Delayed response model, High performance model, Low concurrency model ( Feng et al., 2021 ) CNN Ark shell With high-precision automatic image recognition technology, the accuracy and efficiency of shellfish classification are improved Supervised learning Delayed response model, Lightweight model, High performance model, Low concurrency model ( Kim et al., 2024 ) YOLOv5 Eriocheir sinensis High detection accuracy and excellent recognition effect Supervised learning Delayed response model, Lightweight model, Low concurrency model ( Zhang, Wang, et al., 2023 ) Open in a new tab 2.3. Perspectives AI demonstrates transformative capabilities in aquatic farming and harvesting through core competencies in data-driven modeling, computer vision analytics, deep feature extraction, and multi-modal decision systems. In aquaculture, it enables the transition from conventional to intelligent systems via real-time environmental monitoring, automated disease diagnosis, and dynamic feeding optimization. For harvesting operations, AI facilitates the shift from manual labor to automated efficiency through intelligent capture technologies and AI-driven catch classification. Despite significant promise, shared and domain-specific challenges persist. Common limitations include: compromised data accuracy due to aquatic environmental interference, insufficient model generalizability with limited datasets, and cost barriers for high-precision sensing and computing infrastructure. Domain-specific constraints involve: signal attenuation impeding comprehensive monitoring in large-scale aquaculture, and stability deficits in dynamic harvesting scenarios requiring enhanced model robustness. Collectively, while AI provides foundational support for intelligent transformation, scalable deployment necessitates resolving technical adaptability and cost-efficiency constraints to achieve broader industry adoption. 3. AI-enabled Precision Control in Aquatic Product Processing Conventional processing methods predominantly rely on manual operations and empirical judgments, resulting in inefficiency, imprecision, and limited industrial scalability. Recent AI advancements have catalyzed transformative progress through deep integration of multi-modal monitoring techniques—including spectroscopy, sensor networks, and machine vision. This synergy enhances detection efficiency and accuracy while enabling real-time monitoring of contaminants, freshness indices, and processing parameters. Consequently, it establishes innovative pathways for intelligent, precise, and efficient processing systems, driving the aquatic products industry toward enhanced quality standards and sustainable development ( Fig. 3 A). Fig. 3. Open in a new tab AI applications in aquatic products processing: (A) Comprehensive intelligent innovation across three processing domains; (B) Computer vision and AI-driven mercury detection in fish, with multi-modal spectroscopy-based freshness assessment models ( Maurya et al., 2023 Copyright 2023, Springer; Kashani Zadeh et al., 2023 ); (C) Computational frameworks for fish head positioning and automated freshwater fish cutting ( Zhang, Gong, et al., 2023 ; Peng et al., 2024 ); (D) Dual-AI nutrient enhancement systems for fried fish products ( Sadhu, Banerjee, Lahiri and Chakrabarty, 2020 , Sadhu, Lahiri, Bhattacharjee and Chakrabarty, 2022 Copyright 2020, 2022, Wiley). 3.1. AI Applications in Grading Operations 3.1.1. Contaminant Detection Conventional contaminant detection in aquatic products relies primarily on manual sensory evaluation and laboratory analyses. These methods exhibit three intrinsic limitations: (1) subjective variability in sensory assessment compromises result consistency and reliability; (2) prolonged analytical cycles in laboratory testing impede real-time monitoring for scaled processing; (3) inadequate sensitivity for trace contaminants or microscopic pathogens. Collectively, such constraints hinder efficiency, cost-effectiveness, and precision—critical requirements for modern high-throughput processing systems. AI integration provides innovative solutions to overcome traditional contaminant detection limitations. In spectroscopy, Teng et al. (2023) constructed a surface-enhanced Raman spectroscopy-“one-dimensional Convolutional Neural Network” (1D-CNN) platform for rapid antibiotic residue identification in fish matrices, achieving 100% test accuracy with exceptional specificity. Chen, Liang, et al. (2024) employed liquid chromatography-high resolution mass spectrometry with AIHazardsFinder to classify 32 chemical contaminants. For sensing applications, Chen, You, et al. (2024) developed near-infrared (NIR) carbonized polymer dot sensors integrated with decision tree algorithms, enabling precise Cu 2+ detection in seawater (0.5-80.0 nM) with superior selectivity. Distinctively, Maurya et al. (2023) optimized genetic algorithm-based feature selection with ML classifiers for mercury-contaminated fish identification, achieving 97.1% F1-score. Collectively, these AI-spectroscopy-sensing-vision synergies address fundamental limitations in sensitivity, timeliness, and unknown contaminant detection capabilities. 3.1.2. Freshness Assessment Conventional freshness assessment in aquatic products primarily relies on sensory evaluation and basic instrumental measurements. Trained inspectors manually examine appearance, odor, texture, and coloration characteristics. The quality index method quantifies freshness through full scoring of eye clarity, gill coloration, and muscular elasticity. However, these approaches suffer from susceptibility to subjectivity, inconsistent results, low efficiency, and limited adaptability to diverse species and large-scale processing demands, thereby restricting their applicability in modern operations. Research teams have effectively integrated electronic nose (E-nose), spectroscopy, and AI technologies for fish freshness evaluation. In E-nose applications, Guney and Atasoy (2012) and Atasoy et al. (2015) incorporated ANN, k-nearest neighbors (k-NN), and decision trees into E-nose systems for freshness detection. For spectroscopic approaches, Khoshnoudi-Nia and Moosavi-Nasab (2019) and Kashani Zadeh et al. (2023) coupled spectral techniques with linear/nonlinear regression and ML models ( Fig. 3 B). Additionally, Yasin et al. (2023) and Wu et al. (2022) implemented DL and ML hybrid strategies in this domain. These integrated approaches demonstrate >90% accuracy in freshness assessment. Multi-modal fusion and DL techniques consistently outperform single-method applications, providing robust technical foundations for efficient and objective aquatic product freshness evaluation. For economically significant species including crayfish ( Procambarus clarkii ), shrimp, octopus, red shrimp, and crabs, integrated computer vision and E-nose technologies with AI achieve effective freshness evaluation. Wang, Liu, et al. (2022) developed a CNN-based portable vision system for crayfish freshness assessment. Shao et al. (2018) implemented E-nose with stochastic resonance technology for shrimp freshness analysis. Furthermore, Hu, Zhou, et al. (2020) , Wang et al. (2023) , and Zhu et al. (2019) applied DL to evaluate squid, red shrimp, and crab freshness, respectively. These approaches achieve 86.5%–98.29% accuracy, establishing diverse technological pathways for multi-species freshness assessment. Furthermore, the integration of AI with smart packaging has emerged as a decentralized paradigm for real-time freshness assessment within intelligent processing lines. By utilizing colorimetric sensor arrays ( Cheng et al., 2025 ) and gas-sensitive indicators ( Liang et al., 2024 ), AI-driven platforms can interpret complex colorimetric transitions into precise freshness indices ( Doğan et al., 2024 ). These intelligent packaging solutions, coupled with AI for rapid pattern recognition, bridge the gap between laboratory-based instrumentation and consumer-level quality assurance, significantly advancing the transparency and efficiency of aquatic product quality management ( Li et al., 2023 ). 3.2. AI Applications in Primary Processing 3.2.1. Pre-processing Conventional aquatic product pre-processing—including washing, descaling, and evisceration—relies predominantly on manual operations. This approach exhibits low efficiency, high labor intensity, significant raw material loss, and inadequate standardization, resulting in inconsistent quality and limited industrial scalability. AI technologies offer effective solutions for automated precision pre-processing. Azarmdel et al. (2021) designed a vision-based robotic system for trout processing, integrating machine vision with robotic manipulators to autonomously execute abdominal incision, head removal, evisceration, and cleaning procedures. Nevertheless, AI-enabled pre-processing research remains limited. Further technological development is essential to expand applications for industrial-scale primary processing requirements. 3.2.2. Cutting Operations Conventional aquatic product cutting relies on manual or rudimentary mechanical sectioning, deboning, and despining to achieve product standardization and value enhancement. However, this experience-dependent process exhibits low efficiency, imprecise execution, material waste, and inconsistent quality. Delicate operations like deboning are time-intensive and laborious, with limited adaptability to diverse species and scaled production demands, thereby constraining automation levels. To overcome these limitations, intelligent morphometric assessment has emerged as a crucial precursor to precise primary processing. By integrating computer vision with advanced DL models, AI systems can autonomously quantify key phenotypic traits, such as body dimensions ( Xue et al., 2023 ), mass ( Hamzaoui et al., 2023 ), and fish morphometry ( Saleh et al., 2023 ), enabling the strategic alignment of processing tools with individual biological characteristics. This pre-processing assessment provides a high-fidelity, data-driven foundation for yield optimization, transforming cutting and trimming from generic mechanical operations into personalized, high-precision processes that minimize raw material waste in modern production lines. Building upon this data-driven foundation, recent research has demonstrated the practical integration of AI-led assessment into specific processing tasks. AI technologies are progressively enhancing cutting precision and automation. For fish processing, Zhang, Gong, et al. (2023) , Peng et al. (2024) , and Li et al. (2024) implemented ML coupled with machine vision and laser profiling, achieving >90% mean accuracy in head/tail sectioning while optimizing yield and quality ( Fig. 3 C). In crab processing, Wang et al. (2018) developed a CNN-based method for crab claw localization with 97.67% accuracy. Integrated with k-means clustering, this approach enables precise claw extraction for automated meat extraction systems. 3.3. AI Applications in Advanced Processing Conventional advanced processing of aquatic products—including drying, frying, curing, smoking, and fermentation—imparts distinctive flavors and extended shelf life to meet diversified market demands. However, these methods predominantly rely on manual operations, resulting in inefficiency, inconsistent product quality, and imprecise parameter control that may cause nutritional degradation or suboptimal texture. Furthermore, high-sodium and high-fat formulations raise health concerns while limiting adaptability to scaled production systems and modern consumer preferences for reduced-sodium/low-fat products. To address these constraints, AI technologies enable precise, efficient, and health-oriented advanced processing solutions. Research teams have integrated computer vision and hyperspectral imaging (HSI) with AI models for effective drying process monitoring. Hosseinpour et al. (2012, 2014) implemented computer vision systems with ANN for online monitoring of shrimp drying. For HSI applications, Xu et al. (2022) coupled partial least squares regression with LSSVM, while Wang, Wang, et al. (2024) fused HSI and E-nose data to optimize shrimp drying in real-time. These advances provide robust technological foundations for drying process optimization and quality control, establishing critical groundwork for future multi-modal technology integration. AI technologies demonstrate significant efficacy in frying process optimization, providing scientific methodologies for precise parameter control and enhanced nutritional/quality attributes. Sadhu, Banerjee, Lahiri and Chakrabarty, 2020 , Sadhu, Lahiri, Bhattacharjee and Chakrabarty, 2022 iteratively refined fried fish cooking parameters by integrating ANN with metaheuristic algorithms—including differential evolution, simulated annealing, genetic algorithms, firefly algorithm, and grey wolf optimization—to augment nutritional value ( Fig. 3 D). Complementing this, Chalil George et al. (2022) and Song et al. (2024) employed ANN and ML respectively to optimize frying time-temperature profiles, thereby enhancing product quality. 3.4. Perspectives AI adoption in aquatic product processing exhibits significant sectoral imbalance. Within grading (contaminant detection, freshness assessment), primary processing (cutting operations), and advanced processing (drying, frying), AI demonstrates particular efficacy—achieving high accuracy and practicality across diverse species and multi-modal integration scenarios. Conversely, raw material pre-processing remains underdeveloped, with most studies confined to conceptual designs or isolated operation simulations. Comprehensive intelligent integration of washing, descaling, and evisceration processes has yet to be realized. This technological gap constrains overall processing-chain synergy. Future efforts must prioritize targeted R&D to enable coordinated advancement of AI applications throughout the full processing continuum. 4. Intelligent Transformation in Aquatic Cold Chain Logistics The cold chain constitutes a full engineering process that maintains specific low-temperature environments through integrated refrigeration technologies, precision temperature-control equipment, real-time monitoring systems, and standardized protocols. This ensures product quality stability, safety, and functional integrity. Its core function lies in mitigating biochemical degradation, microbial proliferation, and physical deterioration via seamless low-temperature maintenance and precise regulation, thereby extending shelf life and preserving functional performance. As the critical safeguard for aquatic product quality and safety, cold chain logistics necessitates intelligent advancement. AI technologies—leveraging data processing, pattern recognition, and decision-making capabilities—provide innovative optimization frameworks for transportation, warehousing, and distribution operations ( Fig. 4 ). Fig. 4. Open in a new tab AI integration in aquatic product cold chain logistics: Schematic diagram illustrating AI-enabled optimization across transportation, warehousing, and low-altitude delivery processes ( Hu, Guo, et al., 2023 ; Park et al., 2023 Copyright 2023, Elsevier; Zhang, Saeed, Gao and Hu, 2023 Copyright 2023, IAgrE; Zhao et al., 2025 Copyright 2025, Elsevier). 4.1. AI Applications in Transportation 4.1.1. Demand Forecasting Optimizing demand forecasting algorithms is critical for precise supply chain regulation in aquatic cold chains. Aldahmani et al. (2024) developed a unified demand analysis framework integrating bidirectional long short-term memory with nonlinear autoregressive exogenous models. This effectively captures complex demand dynamics and nonlinear relationships, offering robust solutions for aquatic product demand volatility. Collectively, ML and data mining techniques—through pre-processing and parametric optimization of neural networks, regression analysis, and time-series models—identify multi-dimensional demand patterns driven by seasonal fluctuations, market trends, and supply chain dynamics. These advances significantly enhance forecasting precision, providing implementable solutions for cold chain optimization. 4.1.2. Transport Equipment Optimization and Route Planning In aquatic cold chains, optimizing refrigerated equipment performance and intelligent route planning directly enhance product safety while reducing operational costs. For equipment optimization, Fallmann et al. (2023) developed a dynamic low-order model for small refrigerated vehicles, enabling control algorithm design, soft sensor development, and fault diagnosis. Song et al. (2022) utilized genetic algorithms to optimize electric refrigerated truck systems, significantly extending range and reducing energy consumption. Regarding route planning, Mejjaouli and Babiceanu (2018) implemented a transient virtual machine decision method integrating environmental parameters, cargo location data, and routing protocols for intelligent path planning. Zhang, Tseng, et al. (2019) proposed a ribonucleic acid-inspired ant colony optimization model that reduces transport costs and carbon emission intensity through biomimetic pathfinding. Although these approaches are primarily grounded in traditional mathematical modeling, their parametric optimization and computational efficiency could be further enhanced through ML integration. Expanding beyond these traditional computational models, the integration of digital twins (DT) and Blockchain technology has emerged as a sophisticated frontier for advancing transport equipment intelligence. DT facilitates the creation of high-fidelity virtual replicas of aquatic refrigeration environments, enabling proactive fault diagnosis and predictive maintenance of transport units to minimize the risk of quality degradation. Concurrently, AI-integrated blockchain frameworks establish a decentralized and immutable ledger for multi-source data (MSD) throughout the aquatic product supply chain. This synergy not only ensures the transparency of temperature and quality-related information but also optimizes cold chain logistics through automated, trust-based decision-making, thereby significantly enhancing the overall resilience of aquatic product delivery systems ( Ismail et al., 2023 ). 4.1.3. Temperature Control The high thermal sensitivity of aquatic products necessitates precise temperature regulation during transport to maintain quality. Conventional mechanical cold chains face dual constraints of excessive temperature fluctuations and high energy consumption. AI technologies demonstrate significant potential to address these limitations. Loisel et al. (2022) developed a ML model using neural networks to predict pallet temperature distribution in cold chains, achieving 20%-40% higher accuracy than synthetically trained models. This study provides data-driven decision support for real-time temperature regulation. For multi-temperature refrigerated vehicles, Zou et al. (2023) created an enhanced ANN model integrating LSTM and deep neural networks (DNN). This system enables high-precision temperature estimation via sensor arrays, offering robust monitoring in dynamic environments. Meng et al. (2024) engineered a MSD fusion system for temperature-excursion alerts, attaining 98.6% identification accuracy—significantly outperforming single-source models. This optimizes cost-efficiency in cold chain logistics while enhancing temperature-alert efficacy. Collectively, these advancements demonstrate AI’s capacity to mitigate temperature volatility and reduce energy consumption through predictive modeling, dynamic estimation, and intelligent alert systems, thereby ensuring product safety and quality. 4.2. AI Applications in Warehousing 4.2.1. Cold Storage Control AI technologies including DL, ML, and computer vision are progressively enhancing cold storage operations. For temperature management, Hoang et al. (2021) compared four LSTM variants, identifying convolutional LSTM as optimal for spatiotemporal feature extraction in temperature and power load forecasting. Eze et al. (2024) integrated Pontryagin’s maximum principle with k-means clustering to establish an optimal temperature control model for multi-zone cold chains, supporting quality preservation decisions for perishables like aquatic products. Computer vision enables operational breakthroughs, for example, Liu et al. (2022) developed a CNN-based cargo recognition system achieving 89.92% classification accuracy. This system automates cargo-type identification and robotic gripper selection, reducing manual handling in low-temperature environments while improving efficiency. AI-driven inventory algorithms dynamically adjust stock levels through demand forecasting, significantly increasing turnover rates. While DT and edge computing advance intelligent cold storage, current AI implementations remain fragmented. Future efforts should integrate multi-modal data fusion and cross-system coordination to achieve autonomous decision-making spanning parameter monitoring, energy efficiency, quality control, and inventory management—minimizing human intervention while enhancing operational efficiency and economic returns. 4.2.2. Freeze-Thaw Monitoring Current research demonstrates AI’s potential in food freezing applications, fully categorized into four domains: freezing condition optimization ( Galarce et al., 2025 ), freezing environment control ( Park et al., 2022 ), freezing time prediction ( Ray et al., 2024 ), and quality inspection ( Magdovitz et al., 2022 ). However, studies specifically addressing aquatic products remain scarce. Given the complex characteristics of aquatic freezing processes—including strong ice crystal sensitivity and nonlinear heat transfer—which exhibit high compatibility with AI technologies, future research should prioritize innovative applications of ML and DL in optimizing aquatic product freezing protocols. The AI application framework in food thawing exhibits technical parallels to freezing processes, encompassing four dimensions: thawing condition optimization ( Xia et al., 2026 ), environmental control ( Llave & Erdogdu, 2022 ), time prediction ( Qiao et al., 2024 ), and quality inspection ( Ribault et al., 2021 ). Notably, significant technological gaps persist in aquatic product thawing research. Deep integration of intelligent algorithms with thawing equipment would not only optimize energy consumption but also enable real-time quality monitoring. This dual capability could ensure post-thawing safety and facilitate the transition from empirical control to data-driven intelligent management systems. 4.2.3. Energy Consumption Optimization As a core component of aquatic cold chain warehousing costs, energy optimization critically impacts supply chain economics and environmental sustainability. In intelligent control, Park et al. (2023) developed a deep reinforcement learning-based temperature control framework. This system utilizes IoT-enabled real-time environmental monitoring to reduce energy consumption by 47.64%. Ribault et al. (2021) constructed an ANN-based dynamic electricity pricing response model. By predicting temperature variations via ANN and integrating dynamic programming with metaheuristic algorithms, this approach achieved operational cost reduction. As an innovative extension of classical control theory, Zhu et al. (2023) proposed a dynamically interconnected control strategy that synergistically reduces energy consumption and carbon emissions polynomial-based prediction. Collectively, these approaches—spanning ML optimization, data-driven prediction, and control theory innovation—establish diverse technological pathways for intelligent and low-carbon cold chain warehousing. 4.3. Low-Altitude Delivery AI technologies significantly enhance low-altitude food supply chains. The beetle antennae search algorithm enables real-time collision-free path planning with low computational complexity and robust global search capabilities, balancing energy efficiency and delivery performance. Chen and Shen (2022) developed a risk quantification model integrating k-NN and genetic algorithms to precisely identify traffic congestion risks, optimizing routes for minimized distance and smoothed trajectories to enhance decision reliability. Notably, the perishability of aquatic products imposes stringent demands on delivery timeliness and reliability. Low-altitude drone delivery—with advantages in efficiency, cost control, and environmental sustainability—provides innovative technical solutions for expanding aquatic cold chain logistics scenarios. 4.4. Perspectives AI technologies demonstrate transformative potential across aquatic cold chain logistics by leveraging IoT-enabled real-time sensing, reinforcement learning for dynamic decision-making, and DL for pattern analysis. In transportation, multi-modal data fusion enables precise supply chain regulation, algorithm-optimized equipment and route planning reduce operational costs, and AI-driven temperature control mitigates traditional thermal fluctuations, collectively facilitating a paradigm shift from empirical to data-driven operations. Within warehousing, DL enhances cold storage control efficacy, while research on freeze-thaw monitoring reveals AI’s applicability in optimizing cryogenic conditions and environmental parameters. Energy consumption reduction is achieved through intelligent control systems. For low-altitude delivery, AI supports path planning and dynamic adaptation in food supply chains. Despite these advancements, persistent challenges include superficial AI integration—evidenced by continued reliance on traditional mathematical modeling for route planning—and critical domain-specific gaps in freeze-thaw monitoring, energy optimization, and aquatic product-tailored delivery technologies. Scaling AI implementation necessitates enhanced domain adaptation of food-sector technologies to aquatic contexts and targeted resolution of species-specific technical deficiencies. 5. Digitally-Driven Aquatic Product Marketing Traditional techniques for authenticity verification, sales forecasting, quality control, and personalized recommendations in aquatic product marketing face limitations including inefficiency, subjective biases, and inadequate adaptability. AI technologies offer innovative solutions through multi-source data integration, intelligent feature extraction, and adaptive decision-making capabilities ( Fig. 5 A). Fig. 5. Open in a new tab AI-enabled quality assessment in aquatic product marketing: (A) AI applications spanning sales modalities, technical integration, and future trajectories; (B) Freeze-thaw cycle assessment model for Salmo salar ( Zhang et al., 2024 ) Copyright 2024, Elsevier; (C) Olfactory-based shelf-life quantification system for Eriocheir sinensis ( Zhu et al., 2019 ) Copyright 2019, Wiley. 5.1. Authenticity Verification Authenticity verification has become critical for combating global aquatic product fraud. Malpractices including false origin labeling, species substitution, freeze-thaw adulteration, and chemical preservation threaten consumer rights while compromising product safety and market integrity. ML and DL exhibit significant potential in this domain. Portable NIR spectroscopy coupled with linear/nonlinear ML algorithms enables nutrient assessment, wild/farmed differentiation, geographical origin tracing, and physical state (fresh/frozen) identification to prevent fraud. Currò et al. (2021, 2024) and Varrà et al. (2022) implemented SVM, orthogonal partial least squares-discriminant analysis, and logistic regression with NIR spectroscopy for rapid cephalopod mollusk speciation and geographical provenance discrimination. Chen et al. (2022) compared partial least squares-discriminant analysis, BPNN, and 1D-CNN for salmon-cod admixture detection. The 1D-CNN model achieved optimal 98% classification accuracy. Complementary to NIR spectroscopy, Raman spectroscopy provides a rapid, non-destructive approach for aquatic product authentication. Hu, Huang and Lu (2023) and Ren et al. (2023) achieved over 90% accuracy in fish species verification by integrating Raman spectroscopy with ML models. Furthermore, multi-modal hyperspectral imaging (MSI) coupled with ML enables mislabeling detection, species classification, and freeze-thaw substitution identification in aquatic products. Chauvin et al. (2021) , Sueker et al. (2023) , Xun et al. (2024) , and Tang et al. (2025) employed MSI with advanced algorithms—including simulated annealing, novel ML frameworks, CNN, and dual-branch CNN—to achieve >90% accuracy in species differentiation, textural authentication, mislabeling identification, freeze-thaw substitution detection, and crispness grading. These techniques provide robust technical support for aquatic product safety assurance. Chemical preservation treatments and freeze-thaw adulteration represent prevalent forms of aquatic product fraud, where merchants enhance product appearance or mislabel thawed products as fresh. ML methods detect such practices by analyzing volatile compounds, spectral signatures, electrochemical signals, or specific chemical reactions, thereby ensuring authenticity during sales. Mahata et al. (2024) integrated gas sensors with DNN to discriminate formaldehyde-treated fish from fresh counterparts. Concurrently, Currò et al. (2022) combined NIR spectroscopy with SVM to identify freshness, freeze-thaw status, and hydrogen peroxide treatment in cuttlefish samples. Zhang et al. (2024) developed an ensemble ML approach that achieves 98% accuracy in determining freeze-thaw cycles of high-value Atlantic products ( Fig. 5 B). These advancements demonstrate the considerable potential of AI technologies in aquatic product authentication. 5.2. Sales Forecasting and Quality Control Traditional sales forecasting predominantly relies on historical data, empirical judgments, and basic statistical methods, with limited consideration of market dynamics and external factors. The emergence of e-commerce and community group buying has intensified challenges posed by multi-source heterogeneous data (e.g., consumer behavior, social network metrics). AI technologies enhance sales prediction and process optimization through MSD fusion and advanced algorithms, improving forecast accuracy while enabling real-time data acquisition and analysis. Tian and Wang (2022) developed a backpropagation-long short-term memory-verhulst nested neural network model that elevates sales prediction precision. Furthermore, AI demonstrates extensibility in process optimization: Sykes (2022) implemented an iPad-based computer vision system utilizing Faster R-CNN object detection with TL and data augmentation techniques, achieving 99% recognition accuracy to provide real-time front-end data acquisition for forecasting models. These approaches enhance aquatic product sales prediction accuracy, reduce spoilage from overstocking, enable dynamic strategy adjustments, and improve market adaptability. Growing consumer focus on aquatic product safety and nutritional attributes, driven by consumption upgrades and heightened health awareness, necessitates enhanced quality control. AI technologies provide critical support through innovative applications in supply chain management. Ivorra et al. (2016) , Zhu et al. (2019) , Tian et al. (2020) , and Wijaya et al. (2023) integrated ML/DL with HSI, laser-induced breakdown spectroscopy, and E-nose technologies. This integration achieved 87.2% to 100% accuracy in shelf-life classification, expiration identification, phosphorus quantification, and quality inspection, establishing comprehensive solutions for end-to-end quality control ( Fig. 5 C). 5.3. Personalized Recommendation AI technologies are extensively applied in food recommendation systems for attribute modeling, contextual factor integration, and healthy food suggestions. For attribute modeling, Jia et al. (2022) implemented a multi-view CNN integrated with attention mechanisms, while Zhang, Zhao, et al. (2019) developed a spatially regularized network for multi-ingredient detection to enable personalized recommendations. Regarding contextual factors, Jin et al. (2020) employed DNNs for scenario-based recommendations, and Hamdollahi Oskouei and Hashemzadeh (2023) designed culture-sensitive DL frameworks for cross-cultural precision. In health-focused recommendations, Gautam and Gulhane (2021) proposed multi-constrained PSO, and Perera et al. (2023) created multi-objective recommendation systems to align nutritional profiles with consumer preferences. Although these approaches demonstrate significant efficacy in food recommendations, aquatic product-specific research remains limited. Nevertheless, as aquatic products constitute a critical food subset, their recommendation pipelines—including data pre-processing, model training, and strategy formulation—can leverage methodological parallels from general food recommendation frameworks. Future research should adapt these AI methodologies to aquatic product contexts. 5.4. Perspectives The application of AI technologies in aquatic product sales exemplifies the interdisciplinary convergence of computer science and food science. For authenticity verification, the integration of AI with NIR spectroscopy, Raman spectroscopy, and MSI enables intelligent analysis of multi-source detection data and feature extraction. This significantly enhances precision in species identification, geographical traceability, freeze-thaw adulteration detection, and chemical preservation recognition. Sales forecasting and quality control leverage ML algorithms to conduct in-depth mining of heterogeneous MSD, optimizing predictive decision-making and enabling comprehensive quality management across the supply chain. Although personalized recommendation research remains nascent in aquatic contexts, established methodologies for consumer preference modeling and feature engineering in food science provide transferable frameworks. Nutrition-oriented recommendation systems further expand application boundaries, establishing foundations for precision consumer demand fulfillment. Building upon these conventional ML frameworks, the rapid emergence of Generative AI and large language models (LLMs) is currently catalyzing a more profound paradigm shift in the aquatic product supply chain ( Krupitzer, 2024 ). While standard algorithms excel in structured data analysis, LLMs further bridge the gap in personalized service accuracy by processing unstructured consumer feedback and providing interactive, context-aware dietary recommendations ( Gjorgjevikj et al., 2026 ). These models can seamlessly synthesize complex nutritional profiles with real-time culinary trends to offer bespoke dietary advice and high-quality customer service via intelligent chatbots. Such advancements not only refine the consumer experience but also transform technical quality and safety metrics into accessible, consumer-centric narratives, thereby significantly enhancing engagement and trust in the aquatic product market ( Agrawal et al., 2025 ). Collectively, AI technologies overcome traditional limitations in aquatic product marketing—including inefficiency, subjective biases, and inadequate adaptability—through synergistic integration with multi-spectral detection and data analytics. However, persistent challenges include insufficient exploration of spectral technique-model compatibility and data scarcity for aquatic product personalization. Future research should prioritize optimizing multi-modal data fusion algorithms, developing context-specific recommendation models for aquatic products, and ensuring cost-benefit equilibrium in implementation to maximize AI’s transformative potential for intelligent market advancement. 6. Challenges and future perspectives 6.1. Challenges While the transformative potential of AI in aquatic supply chains is undeniable, the transition from laboratory prototypes to large-scale industrial deployment is constrained by a convergence of technical, operational, and governance challenges. The inherent heterogeneity of dynamic and noisy underwater ecosystems, coupled with the scarcity of high-quality, expert-labeled datasets for niche domains, such as rare species identification or early-stage fish disease recognition, remains a primary bottleneck for training robust ML and DL models. This complexity is further compounded by the variability of data captured via computer vision or multimodal spectroscopy across different sea areas and seasonal cycles, which inherently limits model generalization. Beyond these algorithmic constraints, the practical integration of high-precision AI with hardware faces severe operational hurdles, particularly in processing tasks like fish head positioning or automated filleting that demand sub-second latency and extreme durability. The prohibitive capital expenditure for edge-computing devices resilient to corrosive maritime and high-humidity environments creates a digital divide, hindering the equitable distribution of AI benefits. These barriers are ultimately shadowed by profound trust issues stemming from the “Black-Box” nature of DL; the lack of interpretability hampers regulatory accountability at food safety critical control points, while the risks of proprietary data leaks in decentralized networks continue to challenge the establishment of secure, AI-driven governance frameworks. 6.2. Future Perspectives To bridge the gap toward a fully intelligent ecosystem, future strategic implementation must prioritize advancing model robustness through continuous learning (CL). Shifting from static training to CL paradigms allows AI systems to incrementally acquire knowledge from evolving environmental streams and emerging pathogens without “catastrophic forgetting,” ensuring the long-term reliability of aquatic disease diagnosis. This evolution should be complemented by multi-modal data fusion and edge-cloud synergy, integrating heterogeneous sources like spectroscopy and environmental telemetry to enable real-time, low-latency processing while leveraging cloud-based “Big Data” for logistics optimization. Furthermore, the convergence of AI with IoT and Blockchain offers a promising trajectory for creating tamper-proof architectures, ensuring that freshness data and freeze-thaw cycle assessments are immutable and transparent. Ultimately, fostering cross-disciplinary synergies between food science and management science will dissolve traditional boundaries, transitioning toward a holistic “Farm-to-Table” integration that enhances the overall resilience and sustainability of the global aquatic supply chain. 7. Conclusion This review elucidates the transformative capacity of AI in reshaping aquatic product supply chains, facilitating a critical evolution from conventional labor-intensive practices to sophisticated, data-driven intelligent ecosystems. By synergistically integrating advanced ML and DL architectures across the entire value chain—encompassing precision aquaculture, harvesting, processing, cold-chain logistics, and marketing—AI catalyzes a fundamental paradigm shift toward unprecedented precision, operational transparency, and high efficiency. Despite this promising trajectory, the maturity of AI applications exhibits significant heterogeneity across different operational stages. While high-accuracy models have achieved industrial viability in water quality monitoring, precision feeding strategies, underwater harvesting, and non-destructive freshness assessment, substantial technical bottlenecks remain. Specifically, further research is urgently required to address critical gaps in pre-processing automation, species-specific thawing monitoring, aerial delivery route optimization under complex environmental conditions, and the refinement of personalized recommendation algorithms for consumer engagement. Beyond these technical insights, the impact of this review is profound and twofold: academically, it establishes a robust theoretical framework for interdisciplinary convergence integrating food science, computer science, and management expertise; practically, it serves as a strategic decision-making guide for industry stakeholders aiming to deploy scalable, AI-enabled solutions. Ultimately, this work acts as a vital catalyst for the development of resilient and sustainable aquatic food systems, thereby ensuring global food security while simultaneously fulfilling the escalating consumer demands for premium-quality aquatic products. CRediT authorship contribution statement Xiaonan Fan: Writing – review & editing, Writing – original draft, Validation, Methodology, Funding acquisition, Conceptualization. Jiyu Zou: Writing – original draft, Investigation, Data curation. Yang Liu: Writing – review & editing, Supervision, Methodology, Investigation. Dongmei Li: Validation, Supervision, Conceptualization. Dayong Zhou: Validation, Supervision, Conceptualization. Hai Chi: Methodology, Investigation. Deyang Li: Writing – review & editing, Visualization, Validation, Supervision, Resources, Funding acquisition, Conceptualization. Declaration of competing interest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper. Acknowledgements This work was financially supported by “National Key R&D Program of China (2024YFD2101500)”, “Liaoning Provincial Special Fund Program for Basic Scientific Research Projects in Universities (LJ112410152065)” and “Science and Technology Innovation Special Program of the Liaoning Academy of Agricultural Sciences (2026JC4037, 2025-HBZ-1111)”. Contributor Information Yang Liu, Email: [email protected]. Deyang Li, Email: [email protected]. Data availability No data were generated or analyzed in this study. References Agrawal K., Goktas P., Kumar N., Leung M.F. Artificial intelligence in personalized nutrition and food manufacturing: a comprehensive review of methods, applications, and future directions. Frontiers in Nutrition. 2025;12:1636980. doi: 10.3389/fnut.2025.1636980. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Ahn J.M., Kim J., Kim H., Kim K. Harmful cyanobacterial blooms forecasting based on improved CNN-transformer and temporal fusion transformer. Environmental Technology & Innovation. 2023;32 doi: 10.1016/j.eti.2023.103314. [ DOI ] [ Google Scholar ] Aldahmani E., Alzubi A., Iyiola K. Demand forecasting in supply chain using uni-regression deep approximate forecasting model. Applied Sciences. 2024;14(18):8110. doi: 10.3390/app14188110. [ DOI ] [ Google Scholar ] Ali M.M., Hashim N., Abd Aziz S., Lasekan O. Quality inspection of food and agricultural products using artificial intelligence. Advances in Agricultural and Food Research Journal. 2021;2(2) doi: 10.36877/aafrj.a0000237. [ DOI ] [ Google Scholar ] Atasoy A., Ozsandikcioglu U., Guney S. In 2015 9th International Conference on Electrical and Electronics Engineering (ELECO) 2015. Fish freshness testing with artificial neural networks; pp. 700–704. [ DOI ] [ Google Scholar ] Azarmdel H., Mohtasebi S.S., Jafary A., Behfar H., Rosado Muñoz A. Design and simulation of a vision-based automatic trout fish-processing robot. Applied Sciences. 2021;11(12):5602. doi: 10.3390/app11125602. [ DOI ] [ Google Scholar ] Cao S., Zhao D., Liu X., Sun Y. Real-time robust detector for underwater live crabs based on deep learning. Computers and Electronics in Agriculture. 2020;172 doi: 10.1016/j.compag.2020.105339. [ DOI ] [ Google Scholar ] Cao Y., Lei Q., Wei T., Zhong H. In 2021 International Conference on Electronic Information Engineering and Computer Science (EIECS) 2021. A computer vision program that identifies and classifies fish species; pp. 384–390. [ DOI ] [ Google Scholar ] Chalil George J., Antony S., Ninan G., Kumar K.A., Ravishankar C.N. Artificial neural network models for predicting and optimizing the effect of air-frying time and temperature on physical, textural, sensory, and nutritional quality parameters of fish ball. Journal of Aquatic Food Product Technology. 2022;31(1):35–46. doi: 10.1080/10498850.2021.2008079. [ DOI ] [ Google Scholar ] Chauvin J., Duran R., Tavakolian K., Akhbardeh A., MacKinnon N., Qin J.…Vasefi F. Simulated annealing-based hyperspectral data optimization for fish species classification: Can the number of measured wavelengths Be reduced? Applied Sciences. 2021;11(22):10628. doi: 10.3390/app112210628. [ DOI ] [ Google Scholar ] Chen L., Shen Z. Logistics path decision optimization method of fresh product export cold chain considering transportation risk. Computational Intelligence and Neuroscience. 2022;2022:1–11. doi: 10.1155/2022/8924938. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Chen L., Yang X., Sun C., Wang Y., Xu D., Zhou C. Feed intake prediction model for group fish using the MEA-BP neural network in intensive aquaculture. Information Processing in Agriculture. 2019;7(2):261–271. doi: 10.1016/j.inpa.2019.09.001. [ DOI ] [ Google Scholar ] Chen R., You J., Chen Y., Zhang Z., Sun A., Liu H., Shi X. Broad application prospects near-infrared carbonized polymer dots combined with machine learning for the detection of Cu2+ in seawater and aquatic products. Sensors and Actuators, B: Chemical. 2024;404 doi: 10.1016/j.snb.2023.135245. [ DOI ] [ Google Scholar ] Chen T., Liang W., Zhang X., Wang Y., Lu X., Zhang Y., Zhang Z., You L., Liu X., Zhao C., Xu G. Screening and identification of unknown chemical contaminants in food based on liquid chromatography–high-resolution mass spectrometry and machine learning. Analytica Chimica Acta. 2024;1287 doi: 10.1016/j.aca.2023.342116. [ DOI ] [ PubMed ] [ Google Scholar ] Chen X., Cheng G., Liu S., Meng S., Jiao Y., Zhang W., Liang J., Zhang W., Wang B., Xu X., Xu J. Probing 1D convolutional neural network adapted to near-infrared spectroscopy for efficient classification of mixed fish. Spectrochimica Acta, Part A: Molecular and Biomolecular Spectroscopy. 2022;279 doi: 10.1016/j.saa.2022.121350. [ DOI ] [ PubMed ] [ Google Scholar ] Cheng J., Shen Y., Gu Y., Xiang T., Shen H., Wang Y., Xie Y. Metal-polyphenol Multistage Competitive Coordination System for Colorimetric Monitoring Meat Freshness. Advanced Materials. 2025;37(21):2503246. doi: 10.1002/adma.202503246. [ DOI ] [ PubMed ] [ Google Scholar ] Cromwell J., Turkson C., Dora M., Yamoah F.A. Digital technologies for traceability and transparency in the global fish supply chains: A systematic review and future directions. Marine Policy. 2025;178 doi: 10.1016/j.marpol.2025.106700. [ DOI ] [ Google Scholar ] Currò S., Babbucci M., Carletti P., Fasolato L., Novelli E., Balzan S. The globalized fish industry: Employing DNA-barcoding and NIRS technology to combat counterfeiting and safeguard traditional agro-food products. Food Control. 2024;158 doi: 10.1016/j.foodcont.2023.110246. [ DOI ] [ Google Scholar ] Currò S., Balzan S., Serva L., Boffo L., Ferlito J.C., Novelli E., Fasolato L. Fast and green method to control frauds of geographical origin in traded cuttlefish using a portable infrared reflective instrument. Foods. 2021;10(8):1678. doi: 10.3390/foods10081678. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Currò S., Fasolato L., Serva L., Boffo L., Ferlito J.C., Novelli E., Balzan S. Use of a portable near-infrared tool for rapid on-site inspection of freezing and hydrogen peroxide treatment of cuttlefish (Sepia officinalis) Food Control. 2022;132 doi: 10.1016/j.foodcont.2021.108524. [ DOI ] [ Google Scholar ] De S., Roy, P. C. (Eds.). Advances in energy and sustainability: Select proceedings of INCOM 2024. Springer Nature Singapore. 2024 doi: 10.1007/978-981-97-7308-4. [ DOI ] [ Google Scholar ] Deng Z., Yun Y.-H., Duan N., Wu S. Artificial intelligence algorithms-assisted biosensors in the detection of foodborne pathogenic bacteria: Recent advances and future trends. Trends in Food Science & Technology. 2025;161 doi: 10.1016/j.tifs.2025.105072. [ DOI ] [ Google Scholar ] Doğan V., Evliya M., Kahyaoglu L.N., Kılıç V. On-site colorimetric food spoilage monitoring with smartphone embedded machine learning. Talanta. 2024;266 doi: 10.1016/j.talanta.2023.125021. [ DOI ] [ PubMed ] [ Google Scholar ] Eze J., Duan Y., Eze E., Ramanathan R., Ajmal T. Machine learning-based optimal temperature management model for safety and quality control of perishable food supply chain. Scientific Reports. 2024;14(1):27228. doi: 10.1038/s41598-024-70638-6. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Fallmann M., Poks A., Kozek M. Control-oriented hybrid model of a small-scale refrigerated truck chamber. Applied Thermal Engineering. 2023;220 doi: 10.1016/j.applthermaleng.2022.119719. [ DOI ] [ Google Scholar ] Feng Y., Tao X., Lee E.-J. Classification of shellfish recognition based on improved faster R-CNN framework of deep learning. Mathematical Problems in Engineering. 2021;2021:1–10. doi: 10.1155/2021/1966848. [ DOI ] [ Google Scholar ] Galarce F., Rivera D.R., Pacheco D.R., Caiazzo A., Castillo E. A fast food-freezing temperature estimation framework using optimally located sensors. International Journal of Mechanical Sciences. 2025;299 doi: 10.1016/j.ijmecsci.2025.110374. [ DOI ] [ Google Scholar ] Gautam L.K., Gulhane V.S. In 2021 Fifth International Conference on I-SMAC (Iot in Social, Mobile, Analytics and Cloud) (I-SMAC) 2021. Multi-constraint particle swarm optimization algorithm in diet recommendation for indian elderly persons; pp. 1018–1022. [ DOI ] [ Google Scholar ] Gjorgjevikj A., Martinc M., Cenikj G., Stojanov R., Drole J., Ispirova G., Eftimov T. Large language models in food and nutrition science: Opportunities, challenges, and the case of FoodyLLM. Current Research in Food Science. 2026;12 doi: 10.1016/j.crfs.2026.101351. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] González C., Astudillo C.A., López-Cortés X.A., Maldonado S. Semi-supervised learning for MALDI–TOF mass spectrometry data classification: An application in the salmon industry. Neural Computing and Applications. 2023;35(13):9381–9391. doi: 10.1007/s00521-023-08333-2. [ DOI ] [ Google Scholar ] Guney S., Atasoy A. 2012. An electronic nose system for assessing horse mackerel freshness. In 2012 International Symposium on Innovations in Intelligent Systems and Applications, 1–5. [ DOI ] [ Google Scholar ] Guo J., Dong Y., Lee J.H.W. A real time data driven algal bloom risk forecast system for mariculture management. Marine Pollution Bulletin. 2020;161 doi: 10.1016/j.marpolbul.2020.111731. [ DOI ] [ PubMed ] [ Google Scholar ] Hamdollahi Oskouei S., Hashemzadeh M. FoodRecNet: A comprehensively personalized food recommender system using deep neural networks. Knowledge and Information Systems. 2023;65(9):3753–3775. doi: 10.1007/s10115-023-01897-4. [ DOI ] [ Google Scholar ] Hamzaoui M., Aoueileyine M.O.E., Romdhani L., Bouallegue R. Optimizing XGBoost performance for fish weight prediction through parameter pre-selection. Fishes. 2023;8(10):505. https://www.mdpi.com/2410-3888/8/10/505. [ Google Scholar ] Hao Z., Li C., Zhang H., Ren W. Control and realization of intelligent fishing system based on PLC and fuzzy PID algorithm. Journal of Physics: Conference Series. 2022;2246(1):12042. doi: 10.1088/1742-6596/2246/1/012042. [ DOI ] [ Google Scholar ] Hoang H.M., Akerma M., Mellouli N., Montagner A.L., Leducq D., Delahaye A. Development of deep learning artificial neural networks models to predict temperature and power demand variation for demand response application in cold storage. International Journal of Refrigeration. 2021;131:857–873. doi: 10.1016/j.ijrefrig.2021.07.029. [ DOI ] [ Google Scholar ] Hosseinpour S., Rafiee S., Aghbashlo M., Mohtasebi S.S. Computer vision system (CVS) for In-line monitoring of visual texture kinetics during shrimp (penaeus spp.) drying. Drying Technology. 2014;33(2):238–254. doi: 10.1080/07373937.2014.947513. [ DOI ] [ Google Scholar ] Hosseinpour S., Rafiee S., Mohtasebi S.S., Aghbashlo M. Application of computer vision technique for on-line monitoring of shrimp color changes during drying. Journal of Food Engineering. 2012;115(1):99–114. doi: 10.1016/j.jfoodeng.2012.10.003. [ DOI ] [ Google Scholar ] Hu B., Guo H., Tao X., Zhang Y. Construction of digital twin system for cold chain logistics stereo warehouse. IEEE Access. 2023;11:73850–73862. doi: 10.1109/ACCESS.2023.3295819. [ DOI ] [ Google Scholar ] Hu J., Li D., Duan Q., Han Y., Chen G., Si X. Fish species classification by color, texture and multi-class support vector machine using computer vision. Computers and Electronics in Agriculture. 2012;88:133–140. doi: 10.1016/j.compag.2012.07.008. [ DOI ] [ Google Scholar ] Hu J., Zhou C., Zhao D., Zhang L., Yang G., Chen W. A rapid, low-cost deep learning system to classify squid species and evaluate freshness based on digital images. Fisheries Research. 2020;221 doi: 10.1016/j.fishres.2019.105376. [ DOI ] [ Google Scholar ] Hu W.C., Chen L.B., Yu P.J., Wu M.Y., Huang W.J., Chen K.H. In 2023 5th International Conference on Computer Communication and the Internet (ICCCI) 2023. A deep learning-based white shrimp body length and bait detection methodology for turbid underwater aquaculture environment; pp. 60–65. [ DOI ] [ Google Scholar ] Hu W.C., Wu H.T., Zhang Y.F., Zhang S.H., Lo C.H. Shrimp recognition using ShrimpNet based on convolutional neural network. Journal of Ambient Intelligence and Humanized Computing. 2020 doi: 10.1007/s12652-020-01727-3. [ DOI ] [ Google Scholar ] Hu Y., Huang S.Y., Lu X. A green analytical method for fish species authentication based on raman spectroscopy. Green Analytical Chemistry. 2023;4 doi: 10.1016/j.greeac.2023.100052. [ DOI ] [ Google Scholar ] Huber J., Gossmann A., Stuckenschmidt H. Cluster-based hierarchical demand forecasting for perishable goods. Expert Systems with Applications. 2017;76:140–151. doi: 10.1016/j.eswa.2017.01.022. [ DOI ] [ Google Scholar ] Ismail S., Reza H., Salameh K., Kashani Zadeh H., Vasefi F. Toward an intelligent blockchain IoT-enabled fish supply chain: A review and conceptual framework. Sensors. 2023;23(11):5136. doi: 10.3390/s23115136. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Ivorra E., Sánchez A.J., Verdú S., Barat J.M., Grau R. Shelf life prediction of expired vacuum-packed chilled smoked salmon based on a KNN tissue segmentation method using hyperspectral images. Journal of Food Engineering. 2016;178:110–116. doi: 10.1016/j.jfoodeng.2016.01.008. [ DOI ] [ Google Scholar ] Jang J.S.R. ANFIS: Adaptive-network-based fuzzy inference system. IEEE Transactions on Systems, Man, and Cybernetics. 1993;23(3):665–685. doi: 10.1109/21.256541. [ DOI ] [ Google Scholar ] Javaid M., Haleem A., Khan I.H., Suman R. Understanding the potential applications of artificial intelligence in agriculture sector. Advanced Agrochem. 2023;2(1):15–30. doi: 10.1016/j.aac.2022.10.001. [ DOI ] [ Google Scholar ] Jia N., Chen J., Wang R. An attention-based convolutional neural network for recipe recommendation. Expert Systems with Applications. 2022;201 doi: 10.1016/j.eswa.2022.116979. [ DOI ] [ Google Scholar ] Jiao X., Zhu J., Ye W., Zou H., Yan B., Zhang N., Fan D. Artificial intelligence in smart seafood safety across the supply chains: Recent advances and future prospects. Trends in Food Science & Technology. 2025;163 doi: 10.1016/j.tifs.2025.105161. [ DOI ] [ Google Scholar ] Jin Y., Zhang W., Sun M., Luo X., Wang X. In 2020 IEEE International Conference on Knowledge Graph (ICKG) 2020. Neural restaurant-aware dish recommendation; pp. 599–606. [ DOI ] [ Google Scholar ] Kang H., Zhou H., Wang X., Chen C. Real-time fruit recognition and grasping estimation for robotic apple harvesting. Sensors. 2020;20(19):5670. doi: 10.3390/s20195670. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Kang Z., Zhao Y., Chen L., Guo Y., Mu Q., Wang S. Advances in machine learning and hyperspectral imaging in the food supply chain. Food Engineering Reviews. 2022;14(4):596–616. doi: 10.1007/s12393-022-09322-2. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Kaplan S., Ropelewska E., Günaydın S., Sabancı K., Çetin N. Machine learning and computer vision technology to analyze and discriminate soil samples. Scientific Reports. 2024;14(1):19945. doi: 10.1038/s41598-024-69464-7. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Kariyanna B., Sowjanya M. Unravelling the use of artificial intelligence in management of insect pests. Smart Agricultural Technology. 2024;8 doi: 10.1016/j.atech.2024.100517. [ DOI ] [ Google Scholar ] Kashani Zadeh H., Hardy M., Sueker M., Li Y., Tzouchas A., MacKinnon N.…Elliott C.T. Rapid assessment of fish freshness for multiple supply-chain nodes using multi-mode spectroscopy and fusion-based artificial intelligence. Sensors. 2023;23(11):5149. doi: 10.3390/s23115149. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Khoshnoudi-Nia S., Moosavi-Nasab M. Prediction of various freshness indicators in fish fillets by one multispectral imaging system. Scientific Reports. 2019;9(1):14704. doi: 10.1038/s41598-019-51264-z. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Kim E., Yang S.M., Cha J.E., Jung D.H., Kim H.Y. Deep learning-based phenotype classification of three ark shells: Anadara kagoshimensis, tegillarca granosa, and anadara broughtonii. Frontiers in Marine Science. 2024;11:1356356. doi: 10.3389/fmars.2024.1356356. [ DOI ] [ Google Scholar ] Kollia I., Stevenson J., Kollias S. AI-enabled efficient and safe food supply chain. Electronics. 2021;10(11):1223. doi: 10.3390/electronics10111223. [ DOI ] [ Google Scholar ] Konovalov D.A., Saleh A., Efremova D.B., Domingos J.A., Jerry D.R. 2019. Automatic weight estimation of harvested fish from images. In 2019 Digital Image Computing: Techniques and Applications (DICTA), 1–7. [ DOI ] [ Google Scholar ] Krupitzer C. Generative artificial intelligence in the agri-food value chain-overview, potential, and research challenges. Frontiers in Food Science and Technology. 2024;4:1473357. doi: 10.3389/frfst.2024.1473357. [ DOI ] [ Google Scholar ] Li H., Wang J., Zhang E., Shao Y., Yang L., Yang B., Tan Y., Gao T. Cumulative effects of physical, chemical, and biological measures on algae growth inhibition. Water. 2022;14(6):877. doi: 10.3390/w14060877. [ DOI ] [ Google Scholar ] Li H., Wang Y., Zhang J., Li X., Wang J., Yi S., Li J. Prediction of the freshness of horse mackerel (Trachurus japonicus) using E-nose, E-tongue, and colorimeter based on biochemical indexes analyzed during frozen storage of whole fish. Food Chemistry. 2023;402 doi: 10.1016/j.foodchem.2022.134325. [ DOI ] [ PubMed ] [ Google Scholar ] Li Q., Wen X., Liang S., Sun X., Ma H., Zhang Y., Tan Y., Hong H., Luo Y. Enhancing bighead carp cutting: Chilled storage insights and machine vision-based segmentation algorithm development. Food Chemistry. 2024;450 doi: 10.1016/j.foodchem.2024.139280. [ DOI ] [ PubMed ] [ Google Scholar ] Liang B., Li X., Yang M., Zhang Z., Ren J. Stacking ensemble learning for gas sensor-based detection of salmon freshness and shelf life. Journal of Food Process Engineering. 2024;47(3) doi: 10.1111/jfpe.14593. [ DOI ] [ Google Scholar ] Linaza M.T., Posada J., Bund J., Eisert P., Quartulli M., Döllner J.…Lucat L. Data-driven artificial intelligence applications for sustainable precision agriculture. Agronomy. 2021;11(6):1227. doi: 10.3390/agronomy11061227. [ DOI ] [ Google Scholar ] Liu H., Zhou L., Zhao J., Wang F., Yang J., Liang K., Li Z. Deep-learning-based accurate identification of warehouse goods for robot picking operations. Sustainability. 2022;14(13):7781. doi: 10.3390/su14137781. [ DOI ] [ Google Scholar ] Liu S., Tai H., Ding Q., Li D., Xu L., Wei Y. A hybrid approach of support vector regression with genetic algorithm optimization for aquaculture water quality prediction. Mathematical and Computer Modelling. 2013;58(3–4):458–465. doi: 10.1016/j.mcm.2011.11.021. [ DOI ] [ Google Scholar ] Liu S., Xu L., Li D., Li Q., Jiang Y., Tai H., Zeng L. Prediction of dissolved oxygen content in river crab culture based on least squares support vector regression optimized by improved particle swarm optimization. Computers and Electronics in Agriculture. 2013;95:82–91. doi: 10.1016/j.compag.2013.03.009. [ DOI ] [ Google Scholar ] Llave Y., Erdogdu F. Radio frequency processing and recent advances on thawing and tempering of frozen food products. Critical Reviews in Food Science and Nutrition. 2022;62(3):598–618. doi: 10.1080/10408398.2020.1823815. [ DOI ] [ PubMed ] [ Google Scholar ] Loisel J., Cornuéjols A., Laguerre O., Tardet M., Cagnon D., Duchesne De Lamotte O., Duret S. Machine learning for temperature prediction in food pallet along a cold chain: Comparison between synthetic and experimental training dataset. Journal of Food Engineering. 2022;335 doi: 10.1016/j.jfoodeng.2022.111156. [ DOI ] [ Google Scholar ] Magdovitz B.F., Gummalla S., Garren D., Thippareddi H., Berrang M.E., Love K., Harrison M.A. Analyzing aggregate environmental monitoring data for listeria spp. In frozen food manufacturing environments. Food Control. 2022;135 doi: 10.1016/j.foodcont.2021.108746. [ DOI ] [ Google Scholar ] Mahata B., Acharyya S., Banerji P., Guha P.K. Assessment of fish adulteration using SnO2 nanopetal-based gas sensor and machine learning. Food Chemistry. 2024;438 doi: 10.1016/j.foodchem.2023.138039. [ DOI ] [ PubMed ] [ Google Scholar ] Mainali S. Vol. 208. 2021. Design of an Automatic Fish Processing Line Using Machine Learning. Master’s Theses. [ Google Scholar ] Måløy H., Aamodt A., Misimi E. A spatio-temporal recurrent network for salmon feeding action recognition from underwater videos in aquaculture. Computers and Electronics in Agriculture. 2019;167 doi: 10.1016/j.compag.2019.105087. [ DOI ] [ Google Scholar ] Maurya R., Srivastava A., Srivastava A., Pathak V.K., Dutta M.K. Computer aided detection of mercury heavy metal intoxicated fish: An application of machine vision and artificial intelligence technique. Multimedia Tools and Applications. 2023;82(13):20517–20536. doi: 10.1007/s11042-023-14358-5. [ DOI ] [ Google Scholar ] Mejjaouli S., Babiceanu R.F. Cold supply chain logistics: System optimization for real-time rerouting transportation solutions. Computers in Industry. 2018;95:68–80. doi: 10.1016/j.compind.2017.12.006. [ DOI ] [ Google Scholar ] Meng X., Xie R., Liao J., Shen X., Yang S. A cost-effective over-temperature alarm system for cold chain delivery. Journal of Food Engineering. 2024;368 doi: 10.1016/j.jfoodeng.2023.111914. [ DOI ] [ Google Scholar ] Misra N.N., Dixit Y., Al-Mallahi A., Bhullar M.S., Upadhyay R., Martynenko A. IoT, big data, and artificial intelligence in agriculture and food industry. IEEE Internet of Things Journal. 2022;9(9):6305–6324. doi: 10.1109/JIOT.2020.2998584. [ DOI ] [ Google Scholar ] Pallathadka H., Jawarneh M., Sammy F., Garchar V., Sanchez D.T., Naved M. In 2022 2nd International Conference on Advance Computing and Innovative Technologies in Engineering (ICACITE) 2022. A review of using artificial intelligence and machine learning in food and agriculture industry; pp. 2215–2218. [ DOI ] [ Google Scholar ] Park B.R., Choi Y.J., Choi E.J., Moon J.W. Adaptive control algorithm with a retraining technique to predict the optimal amount of chilled water in a data center cooling system. Journal of Building Engineering. 2022;50 doi: 10.1016/j.jobe.2022.104167. [ DOI ] [ Google Scholar ] Park J., Park T., Lee D., Shin J., Kim K., Jung J., Cha Y. Generalizable deep learning forecasting of harmful algal blooms using transfer learning across river systems. Ecological Informatics. 2025;103481 doi: 10.1016/j.ecoinf.2025.103481. [ DOI ] [ Google Scholar ] Park J.-W., Ju Y.-M., Kim Y.-G., Kim H.-S. 50% reduction in energy consumption in an actual cold storage facility using a deep reinforcement learning-based control algorithm. Applied Energy. 2023;352 doi: 10.1016/j.apenergy.2023.121996. [ DOI ] [ Google Scholar ] Peng X., Chen Y., Fu D., Jiang Y., Hu Z. Study on visual localization and evaluation of automatic freshwater fish cutting system based on deep learning framework. International Journal of Food Properties. 2024;27(1):516–531. doi: 10.1080/10942912.2024.2330503. [ DOI ] [ Google Scholar ] Perera Y., Karannagoda R., Weiman D., Fernando S. In 2023 8th International Conference on Information Technology Research (ICITR) 2023. Multiple objective optimization based dietary recommender system; pp. 1–6. [ DOI ] [ Google Scholar ] Pourkabirian A., Kooshki F., Anisi M.H., Jindal A. An accurate RSS/AoA-based localization method for internet of underwater things. Ad Hoc Networks. 2023;145 doi: 10.1016/j.adhoc.2023.103177. [ DOI ] [ Google Scholar ] Qiao J., Zhang M., Wang D., Mujumdar A.S., Chu C. AI-based R&D for frozen and thawed meat: Research progress and future prospects. Comprehensive Reviews in Food Science and Food Safety. 2024;23(5) doi: 10.1111/1541-4337.70016. [ DOI ] [ PubMed ] [ Google Scholar ] Qiao X., Bao J., Zeng L., Zou J., Li D. An automatic active contour method for sea cucumber segmentation in natural underwater environments. Computers and Electronics in Agriculture. 2017;135:134–142. doi: 10.1016/j.compag.2017.02.008. [ DOI ] [ Google Scholar ] Ray A., Minz P.S., Sinha C. Framework for accurate estimation of freezing time and convective heat transfer coefficient for freezing of a food product in domestic refrigerator: a numerical and simulation modeling approach. Multiscale and Multidisciplinary Modeling, Experiments and Design. 2024;7(6):5481–5498. doi: 10.1007/s41939-024-00533-0. [ DOI ] [ Google Scholar ] Ren L., Tian Y., Yang X., Wang Q., Wang L., Geng X., Wang K., Du Z., Li Y., Lin H. Rapid identification of fish species by laser-induced breakdown spectroscopy and raman spectroscopy coupled with machine learning methods. Food Chemistry. 2023;400 doi: 10.1016/j.foodchem.2022.134043. [ DOI ] [ PubMed ] [ Google Scholar ] Ribault A., Vercraene S., Henry S., Ouzrout Y. Economic optimisation of cold production: A matheuristic with artificial neural network approach. International Journal of Production Research. 2021;59(22):6941–6962. doi: 10.1080/00207543.2020.1831705. [ DOI ] [ Google Scholar ] Sadhu T., Banerjee I., Lahiri S.K., Chakrabarty J. Modeling and optimization of cooking process parameters to improve the nutritional profile of fried fish by robust hybrid artificial intelligence approach. Journal of Food Process Engineering. 2020;43(9) doi: 10.1111/jfpe.13478. [ DOI ] [ Google Scholar ] Sadhu T., Lahiri S.K., Bhattacharjee A., Chakrabarty J. Application of artificial neural network with metaheuristic optimization for improving the nutritive value of fried fish. Journal of Food Processing and Preservation. 2022;46(12) doi: 10.1111/jfpp.17190. [ DOI ] [ Google Scholar ] Saleh A., Jones D., Jerry D., Azghadi M.R. MFLD-net: a lightweight deep learning network for fish morphometry using landmark detection. Aquatic Ecology. 2023;57(4):913–931. doi: 10.1007/s10452-023-10044-8. [ DOI ] [ Google Scholar ] Shao C., Zheng H., Zhou Z., Li J., Lou X., Hui G., Zhao Z. Ridgetail white prawn (exopalaemon carinicauda) K value predicting method by using electronic nose combined with non-linear data analysis model. Food Analytical Methods. 2018;11(11):3121–3129. doi: 10.1007/s12161-018-1297-8. [ DOI ] [ Google Scholar ] Shen H.-T., Pan X.-D., Cao L.-N., Han J.-L. Distribution and health risk assessment of multiple elements in canned aquatic products from southeast China. Food Control. 2025;175 doi: 10.1016/j.foodcont.2025.111270. [ DOI ] [ Google Scholar ] Song G., Zeng M., Chen S., Lyu Z., Jiang N., Wang D., Yuan T., Li L., Mei G., Shen Q., Gong J. Exploring molecular mechanisms underlying changes in lipid fingerprinting of salmon (salmo salar) during air frying integrating machine learning-guided REIMS and lipidomics analysis. Food Chemistry. 2024;460 doi: 10.1016/j.foodchem.2024.140770. [ DOI ] [ PubMed ] [ Google Scholar ] Song H., Cai M., Cen J., Xu C., Zeng Q. Research on energy saving optimization method of electric refrigerated truck based on genetic algorithm. International Journal of Refrigeration. 2022;137:62–69. doi: 10.1016/j.ijrefrig.2022.02.003. [ DOI ] [ Google Scholar ] Sueker M., Daghighi A., Akhbardeh A., MacKinnon N., Bearman G., Baek I.…Kashani Zadeh H. A novel machine-learning framework based on a hierarchy of dispute models for the identification of fish species using multi-mode spectroscopy. Sensors. 2023;23(22):9062. doi: 10.3390/s23229062. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Sykes E.R. A deep learning computer vision iPad application for sales rep optimization in the field. Visual Computer. 2022;38(2):729–748. doi: 10.1007/s00371-020-02047-5. [ DOI ] [ Google Scholar ] Tang S., Zhang L., Tian X., Zheng M., Zhang H., Zhong N. Synergizing meat science and interpretable AI: Quantifying crispness gradients for quality authentication of tilapia fillet processing. Food Chemistry. 2025;484 doi: 10.1016/j.foodchem.2025.144252. [ DOI ] [ PubMed ] [ Google Scholar ] Teng Y., Wang Z., Zuo S., Li X., Chen Y. Identification of antibiotic residues in aquatic products with surface-enhanced raman scattering powered by 1-D convolutional neural networks. Spectrochimica Acta, Part A: Molecular and Biomolecular Spectroscopy. 2023;289 doi: 10.1016/j.saa.2022.122195. [ DOI ] [ PubMed ] [ Google Scholar ] Tian L., Wang X. A dynamic prediction neural network model of cross-border e-commerce sales for virtual community knowledge sharing. Computational Intelligence and Neuroscience. 2022;2022:1–11. doi: 10.1155/2022/2529372. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Tian Y., Chen Q., Lin Y., Lu Y., Li Y., Lin H. Quantitative determination of phosphorus in seafood using laser-induced breakdown spectroscopy combined with machine learning. Spectrochimica Acta, Part B: Atomic Spectroscopy. 2020;175 doi: 10.1016/j.sab.2020.106027. [ DOI ] [ Google Scholar ] Varrà M.O., Ghidini S., Fabrile M.P., Ianieri A., Zanardi E. Country of origin label monitoring of musky and common octopuses (eledone spp. And octopus vulgaris) by means of a portable near-infrared spectroscopic device. Food Control. 2022;138 doi: 10.1016/j.foodcont.2022.109052. [ DOI ] [ Google Scholar ] Waleed A., Medhat H., Esmail M., Osama K., Samy R., Ghanim T.M. In 2019 14th International Conference on Computer Engineering and Systems (ICCES) 2019. Automatic recognition of fish diseases in fish farms; pp. 201–206. [ DOI ] [ Google Scholar ] Wang C., Liu Y., Xia Z., Wang Q., Duan S., Gong Z., Chen J. Convolutional neural network-based portable computer vision system for freshness assessment of crayfish (Prokaryophyllus clarkii) Journal of Food Science. 2022;87(12):5330–5339. doi: 10.1111/1750-3841.16377. [ DOI ] [ PubMed ] [ Google Scholar ] Wang D., Vinson R., Holmes M., Seibel G. Convolutional neural network guided blue crab knuckle detection for autonomous crab meat picking machine. Optical Engineering. 2018;57(4):1. doi: 10.1117/1.OE.57.4.043103. [ DOI ] [ Google Scholar ] Wang J., Wang W., Xu W., An H., Ma Q., Sun J., Wang J. Fusing hyperspectral imaging and electronic nose data to predict moisture content in penaeus vannamei during solar drying. Frontiers in Nutrition. 2024;11:1220131. doi: 10.3389/fnut.2024.1220131. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Wang K., Zhang C., Wang R., Ding X. Quality non-destructive diagnosis of red shrimp based on image processing. Journal of Food Engineering. 2023;357 doi: 10.1016/j.jfoodeng.2023.111648. [ DOI ] [ Google Scholar ] Wang S., Guo J., Guo S., Fu Q., Xu J. In 2022 IEEE International Conference on Mechatronics and Automation (ICMA) 2022. Study on real-time recognition of underwater live shrimp by the spherical amphibious robot based on deep learning; pp. 917–922. [ DOI ] [ Google Scholar ] Wang Y., Gu H.W., Yin X.L., Geng T., Long W., Fu H., She Y. Deep leaning in food safety and authenticity detection: An integrative review and future prospects. Trends in Food Science & Technology. 2024;146 doi: 10.1016/j.tifs.2024.104396. [ DOI ] [ Google Scholar ] Wei H., Peng D., Zhu X., Dongdong W. In 2016 2nd IEEE International Conference on Computer and Communications (ICCC) 2016. A target tracking algorithm for vision based sea cucumber capture; pp. 401–404. [ DOI ] [ Google Scholar ] Wei X., Zhang M., Chen K., Huang M., Mujumdar A.S., Yang C. Intelligent detection and control of quality deterioration of fresh aquatic products in the supply chain: A review. Computers and Electronics in Agriculture. 2024;218 doi: 10.1016/j.compag.2024.108720. [ DOI ] [ Google Scholar ] Wijaya D.R., Syarwan N.F., Nugraha M.A., Ananda D., Fahrudin T., Handayani R. Seafood quality detection using electronic nose and machine learning algorithms with hyperparameter optimization. IEEE Access. 2023;11:62484–62495. doi: 10.1109/ACCESS.2023.3286980. [ DOI ] [ Google Scholar ] Wu T., Lu J., Zou J., Chen N., Yang L. Accurate prediction of salmon freshness under temperature fluctuations using the convolutional neural network long short-term memory model. Journal of Food Engineering. 2022;334 doi: 10.1016/j.jfoodeng.2022.111171. [ DOI ] [ Google Scholar ] Xia Q., Yan S., Huang M., Chen K., Huang J. Advances in Freezing and Thawing Meat: From Physical Principles to Artificial Intelligence. Foods. 2026;15(2):396. doi: 10.3390/foods15020396. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Xu W., Zhang F., Wang J., Ma Q., Sun J., Tang Y., Wang J., Wang W. Real-time monitoring of the quality changes in shrimp (penaeus vannamei) with hyperspectral imaging technology during hot air drying. Foods. 2022;11(20):3179. doi: 10.3390/foods11203179. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Xue Y., Bastiaansen J.W., Khan H.A., Komen H. An analytical framework to predict slaughter traits from images in fish. Aquaculture. 2023;566 doi: 10.1016/j.aquaculture.2022.739175. [ DOI ] [ Google Scholar ] Xun Z., Wang X., Xue H., Zhang Q., Yang W., Zhang H., Li M., Jia S., Qu J., Wang X. Deep machine learning identified fish flesh using multispectral imaging. Current Research in Food Science. 2024;9 doi: 10.1016/j.crfs.2024.100784. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Yang C., Guo Z., Fernandes Barbin D., Dai Z., Watson N., Povey M., Zou X. Hyperspectral imaging and deep learning for quality and safety inspection of fruits and vegetables: A review. Journal of Agricultural and Food Chemistry. 2025;73(17):10019–10035. doi: 10.1021/acs.jafc.4c11492. [ DOI ] [ PubMed ] [ Google Scholar ] Yang M., Wang W., Gao Q., Zhao C., Li C., Yang X., Li J., Li X., Cui J., Zhang L., Ji Y., Geng S. Automatic identification of harmful algae based on multiple convolutional neural networks and transfer learning. Environmental Science and Pollution Research. 2022;30(6):15311–15324. doi: 10.1007/s11356-022-23280-6. [ DOI ] [ PubMed ] [ Google Scholar ] Yasin E.T., Ozkan I.A., Koklu M. Detection of fish freshness using artificial intelligence methods. European Food Research and Technology. 2023;249(8):1979–1990. doi: 10.1007/s00217-023-04271-4. [ DOI ] [ Google Scholar ] Yuan Y., Ji Z., Fan Y., Xu Q., Shi C., Lyu J., Ertbjerg P. Deep learning-assisted fluorescence spectroscopy for food quality and safety analysis. Trends in Food Science & Technology. 2025;156 doi: 10.1016/j.tifs.2024.104821. [ DOI ] [ Google Scholar ] Yumnam M., Gopalakrishnan K., Dhua S., Srivastava Y., Mishra P. A comprehensive review on smartphone-based sensor for fish spoilage analysis: Applications and limitations. Food and Bioprocess Technology. 2024;17(12):4575–4597. doi: 10.1007/s11947-024-03391-3. [ DOI ] [ Google Scholar ] Zhang D., Lillywhite K.D., Lee D.-J., Tippetts B.J. Automatic shrimp shape grading using evolution constructed features. Computers and Electronics in Agriculture. 2014;100:116–122. doi: 10.1016/j.compag.2013.11.009. [ DOI ] [ Google Scholar ] Zhang D., Zhang L., Cong B., Hu J., Zhang X. Multi-frequency bioimpedance combined with machine learning for frozen-thawed quality evaluation of atlantic salmon. Journal of Food Engineering. 2024;382 doi: 10.1016/j.jfoodeng.2024.112195. [ DOI ] [ Google Scholar ] Zhang H., Yu F., Sun J., Shen X., Li K. Deep learning for sea cucumber detection using stochastic gradient descent algorithm. European Journal of Remote Sensing. 2020;53(sup1):53–62. doi: 10.1080/22797254.2020.1715265. [ DOI ] [ Google Scholar ] Zhang J., Wang S., Zhang S., Li J., Sun Y. Research on target detection and recognition algorithm of eriocheir sinensis carapace. Multimedia Tools and Applications. 2023;82(27):42527–42543. doi: 10.1007/s11042-023-15228-w. [ DOI ] [ Google Scholar ] Zhang L., Li W., Liu C., Zhou X., Duan Q. Automatic fish counting method using image density grading and local regression. Computers and Electronics in Agriculture. 2020;179 doi: 10.1016/j.compag.2020.105844. [ DOI ] [ Google Scholar ] Zhang L., Saeed R., Gao Q., Hu J. Information fusion enabled system for monitoring the vitality of live crabs during transportation. Biosystems Engineering. 2023;235:50–68. doi: 10.1016/j.biosystemseng.2023.09.007. [ DOI ] [ Google Scholar ] Zhang L., Zhao J., Li S., Shi B., Duan L.Y. In 2019 IEEE International Conference on Multimedia and Expo (ICME) 2019. From market to dish: Multi-ingredient image recognition for personalized recipe recommendation; pp. 1252–1257. [ DOI ] [ Google Scholar ] Zhang L.Y., Tseng M.L., Wang C.H., Xiao C., Fei T. Low-carbon cold chain logistics using ribonucleic acid-ant colony optimization algorithm. Journal of Cleaner Production. 2019;233:169–180. doi: 10.1016/j.jclepro.2019.05.306. [ DOI ] [ Google Scholar ] Zhang X., Gong Z., Liang X., Sun W., Ma J., Wang H. Line laser scanning combined with machine learning for fish head cutting position identification. Foods. 2023;12(24):4518. doi: 10.3390/foods12244518. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Zhang Y., Yue J., Song A., Jia S., Li Z. A high-similarity shellfish recognition method based on convolutional neural network. Information Processing in Agriculture. 2023;10(2):149–163. doi: 10.1016/j.inpa.2022.05.009. [ DOI ] [ Google Scholar ] Zhao F., Zhang C., Abujar S. In 2023 IEEE International Conference on Multimedia and Expo Workshops (ICMEW) IEEE; 2023. A Multimodal Approach for Evaluating Algal Bloom Severity Using Deep Learning; pp. 506–509. [ DOI ] [ Google Scholar ] Zhao S., Jin F.F., Stuecker M.F., Thompson P.R., Kug J.-S., McPhaden M.J.…Cai W. Explainable el niño predictability from climate mode interactions. Nature. 2024;630(8018):891–898. doi: 10.1038/s41586-024-07534-6. [ DOI ] [ PubMed ] [ Google Scholar ] Zhao S., Zhang S., Liu J., Wang H., Zhu J., Li D., Zhao R. Application of machine learning in intelligent fish aquaculture: A review. Aquaculture. 2021;540 doi: 10.1016/j.aquaculture.2021.736724. [ DOI ] [ Google Scholar ] Zhao Y., Zhang M., Mujumdar A.S., Adhikari B., Rui L. AI-Based Low-Altitude Delivery Fresh Food Supply Chain: Research Progress and Trends. Trends in Food Science & Technology. 2025;161 doi: 10.1016/j.tifs.2025.105056. [ DOI ] [ Google Scholar ] Zhou C., Lin K., Xu D., Chen L., Guo Q., Sun C., Yang X. Near infrared computer vision and neuro-fuzzy model-based feeding decision system for fish in aquaculture. Computers and Electronics in Agriculture. 2018;146:114–124. doi: 10.1016/j.compag.2018.02.006. [ DOI ] [ Google Scholar ] Zhou C., Xu D., Chen L., Zhang S., Sun C., Yang X., Wang Y. Evaluation of fish feeding intensity in aquaculture using a convolutional neural network and machine vision. Aquaculture. 2019;507:457–465. doi: 10.1016/j.aquaculture.2019.04.056. [ DOI ] [ Google Scholar ] Zhou Z., Zahid U., Majeed Y.…Nisha Mustafa. Advancement in artificial intelligence for on-farm fruit sorting and transportation. Frontiers in Plant Science. 2023;14:1082860. doi: 10.3389/fpls.2023.1082860. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Zhu P., Zhang Y., Chou Y., Gu Y. Recognition of the storage life of mitten crab by a machine olfactory system with deep learning. Journal of Food Process Engineering. 2019;42(7) doi: 10.1111/jfpe.13095. [ DOI ] [ Google Scholar ] Zhu Z., Liu X., Zhao S., Shan X., Chen A., Yu J., Liu B. Energy saving and carbon emission reduction potential for cold store with new dynamic linkage control strategy. International Journal of Refrigeration. 2023;154:43–55. doi: 10.1016/j.ijrefrig.2023.07.001. [ DOI ] [ Google Scholar ] Zou Y., Wu J., Wang X., Morales K., Liu G., Manzardo A. An improved artificial neural network using multi-source data to estimate food temperature during multi-temperature delivery. Journal of Food Engineering. 2023;351 doi: 10.1016/j.jfoodeng.2023.111518. [ DOI ] [ Google Scholar ] Associated Data This section collects any data citations, data availability statements, or supplementary materials included in this article. Data Availability Statement No data were generated or analyzed in this study. Articles from Food Chemistry: X are provided here courtesy of Elsevier ACTIONS View on publisher site PDF (9.6 MB) Cite Collections Permalink PERMALINK Copy RESOURCES Similar articles Cited by other articles Links to NCBI Databases Cite Copy Download .nbib .nbib Format: AMA APA MLA NLM Add to Collections Create a new collection Add to an existing collection Name your collection * Choose a collection Unable to load your collection due to an error Please try again Add Cancel Follow NCBI NCBI on X (formerly known as Twitter) NCBI on Facebook NCBI on LinkedIn NCBI on GitHub NCBI RSS feed Connect with NLM NLM on X (formerly known as Twitter) NLM on Facebook NLM on YouTube National Library of Medicine 8600 Rockville Pike Bethesda, MD 20894 Web Policies FOIA HHS Vulnerability Disclosure Help Accessibility Careers NLM NIH HHS USA.gov Back to Top