Search Results (3,540)

Illegal, unreported, and unregulated fishing threatens marine ecosystem health and sustainable fisheries management, highlighting the need for reliable fishing-vessel behavior recognition from Automatic Identification System (AIS) trajectories. However, AIS-derived operational states often exhibit overlapping motion patterns, particularly between Underway and Fishing and between Anchored and Moored. This study proposes FishFormer, a local–global spatio-temporal deep learning framework designed for recognizing four AIS-status-derived fishing-vessel operational states: Underway, Fishing, Anchored, and Moored. FishFormer integrates dual-stream spatio-temporal attention, local–global feature fusion, and feed-forward feature enhancement to capture long-range trajectory dependencies, local motion variations, and heterogeneous kinematic features. Experiments on 8139 real-world AIS trajectory segments from U.S. coastal waters show that FishFormer achieves 96.63% overall accuracy and an F1-score of 0.9661. Compared with seven baseline models under a unified experimental protocol, FishFormer shows superior recognition performance, while ablation, confusion-matrix, and robustness analyses further verify the effectiveness of the proposed modules and their contribution to reducing errors among similar behavior states. These results indicate that local–global spatio-temporal learning improves AIS-based operational-state recognition and can provide a behavioral information layer for fishing-vessel activity monitoring and fishery management decision support. Full article

The global sugar industry is facing increasing challenges due to climate variability, sustainability requirements, and the need for improved operational efficiency. These pressures are driving the search for advanced technological solutions to enhance productivity and resource management. Artificial intelligence (AI) has already demonstrated significant potential across various agricultural sectors; however, a comprehensive evaluation of AI applications across the entire sugar industry value chain from crop cultivation to industrial processing and supply chain management remains limited. This review provides a detailed assessment of the current state of AI and internet of things (IoT) implementation in the sugar beet industry. It examines key applications, including precision agriculture for sugarcane and sugar beet cultivation, intelligent monitoring systems for early disease detection, and AI-driven decision support tools for resource optimization. In addition, the study explores the role of AI in sugar manufacturing processes, where machine learning and data-driven models are used to optimize milling operations, improve product quality control, and enable predictive maintenance of industrial equipment. AI technologies are also shown to enhance supply chain efficiency through improved demand forecasting, logistics optimization, and real-time data analytics. Monitoring volatile organic compounds (VOCs) is becoming increasingly important in sugar beet and sugarcane storage. Microbial activity during storage and fermentation can release VOCs such as ethanol, which act as early indicators of crop degradation and spoilage. Detecting these gases using modern gas sensors enables continuous monitoring of storage conditions and crop health. When sensor data is integrated with AI and IoT systems, it can be analyzed in real time to identify early signs of microbial activity, improve storage management, and optimize processing decisions. Such intelligent monitoring systems have the potential to reduce losses and enhance overall efficiency in the sugar production chain. Full article

Background and Objectives: Despite recent advancement, neovascular age-related macular degeneration (nAMD) remains a leading cause of irreversible vision loss. Undertreatment, fewer anti-VEGF injections and longer intervals than in clinical trials have been associated with sub-optimal visual outcomes. Visit-based regimens (Treat-and-Extend, PRN) may permit intervals of unrecognized retinal fluid between office visits. A home OCT system with near-daily self-imaging provides frequent structural retinal information between office visits that can support early detection of persistent or recurring fluid. The objective was to evaluate the duration and magnitude of fluid exposure between standard care visits and estimate the potential to shorten that exposure. Materials andMethods: Ad hoc analysis of three cohorts of treatment naïve and experienced nAMD eyes managed by standard care while participating in observational studies of the home OCT system, with treating physicians masked to home OCT data. AI-based analysis of fluid volume, rate of change and time of fluid onset was performed. Results: Data from 209 participants, mean age 76.4 years, 53% female, who performed 10,110 scans (6.0 scans/week) were analyzed. An amount of 119 eligible eyes provided data from 185 standard care intervals. Persistent or recurring fluid was identified in 121 (65%) intervals, on average 32 days prior to the next office visit. Of these, 84 (69%) had potential visit advancement within labeled minimal treatment intervals of 19 days. Mean fluid volume at the earliest possible notification was 26 nL and recurrence rate averaged 4.4 nL/day. Conclusions: A substantial proportion of patients experience unrecognized disease activity between visits. Home OCT monitoring provides adjunctive information to support early detection of fluid and may facilitate timely clinical evaluation. In this context, such monitoring may be considered reasonable and necessary to inform management of nAMD within established standards of care, while not replacing clinician-directed diagnosis or treatment decisions. Full article

Artificial Intelligence (AI) has become a key enabler in soil science and agriculture, supporting advanced modeling, monitoring, and decision-making processes. This systematic review synthesizes recent developments in AI-based soil characterization and agricultural applications, with emphasis on soil physicochemical properties, digital soil mapping, irrigation management, and crop yield prediction. Following the PRISMA 2020 framework, a structured search of the Scopus database identified 196 eligible studies published between 2018 and 2026. The reviewed literature reveals a clear transition toward data-driven approaches, with machine learning and deep learning models dominating recent research. Random Forest, Support Vector Machines, gradient boosting methods, artificial neural networks, Convolutional Neural Networks, and Long Short-Term Memory architectures were the most frequently reported techniques. The primary data sources included in situ sensors, laboratory measurements, remote sensing imagery, and environmental covariates, often integrated through multi-source data fusion frameworks. The results indicate that tree-based ensemble models provide robust performance across diverse soil properties, whereas deep learning models are particularly effective for spatiotemporal prediction and remote sensing applications. AI-driven systems are increasingly used to support precision agriculture through irrigation optimization, crop yield forecasting, digital soil mapping, and soil health monitoring. However, challenges remain regarding data quality and availability, model transferability across regions, and the limited interpretability of complex models. The findings highlight current research trends, methodological challenges, and future opportunities for the development of reliable and scalable AI-driven soil and agricultural systems. Full article

The implementation of an artificial intelligence (AI)-based system for monitoring, forecasting, and learner performance support has been intensified by the rapid expansion of online education systems. Existing online educational platforms completely rely on learning analytics and machine learning to customize content delivery. On the other hand, these platforms fundamentally focus on behavioral and cognitive indicators, whereas the integration of affective computing into learning analytics and adaptive decision-making processes is lacking. During the learning process, emotions like engagement, boredom, and confusion play a vital role. Nonetheless, the integration of adaptive online learning systems is still fragmented and underdeveloped. The latest progress in affective computing and multimodal sensing technologies allow for the inference of the affective state of learners in real-time, which creates a range of potential opportunities to create emotionally sensitive learning spaces. Despite technological innovations, the existing studies do not have a conceptual framework that is unified, design-oriented, and clearly incorporates affective computing with AI-based learning analytics to inform real-time pedagogical adaptation. To address this gap, this study introduces a design-oriented conceptual framework for AI-based online education systems that incorporate real-time affective computing. This conceptual framework combines the theoretical foundation of learning analytics, affective computing, and adaptive learning systems. The suggested framework offers a clear and scalable basis of online learning environments that are affective-aware by offering a clear framework of development, assessment, and consequent empirical validation. Full article

Radiation monitoring with SRAM-based FPGAs traditionally relies on offset-histogram analysis, which requires a chip-specific calibration campaign at an accelerator before multiple-cell upsets (MCUs) can be discriminated from coincident single-cell upsets (SCUs). The cost and complexity of such calibration restrict the approach to dedicated, beam-test-funded programs. We propose an AI-based on-board classifier that achieves MCU/SCU discrimination directly, without any chip-specific calibration. A lightweight Multi-Layer Perceptron (MLP), trained entirely on synthetic data covering five representative bit-interleaving layouts, is integrated on an AMD Artix-7 XC7A200T FPGA together with per-detection-element telemetry aggregation. The classifier achieves F1 = 0.92–0.97 on structured BRAM layouts when per-chip calibration data are available (calibrated ceiling) and, without any chip-specific calibration, retains F1 up to 0.81 ± 0.02 (held-out, mean over five seeds) on previously unseen layouts with near-perfect recall. A sensitivity analysis across a $20\times$ range of SEU rates and a $4\times$ range of MCU fractions confirms the robustness of the proposed approach. A feature-ablation study identifies an indispensable feature subset, while a comparative evaluation of four alternative classifier architectures (decision tree, support vector machine (SVM), two MLP variants) establishes the reference MLP as the optimal choice. Post-implementation results on the Artix-7 200T show that the MLP-enhanced and calibrated-histogram designs occupy nearly identical FPGA footprints, reframing the choice between them as an operational decision driven by calibration availability rather than by hardware cost. Full article

Background and Objectives: As Romanian health systems deploy artificial intelligence (AI), uptake depends on navigating the EU AI Act, GDPR, the Medical Device Regulation (MDR), and national rules. We measured AI legislation literacy, governance readiness, and adoption intentions among Romanian healthcare professionals, identified implementation phenotypes, and tested whether confidence mediates the literacy–adoption link. Materials and Methods: In a multicenter cross-sectional survey (N = 109), participants completed a 20-item AI Legislation Literacy Index (0–20) plus scales rated form one to five measuring legislative confidence, adoption intention, readiness, trust, and perceived compliance burden. We used PCA and k-means clustering, multivariable logistic regression for high adoption intention (≥4), and covariate-adjusted mediation (5000 bootstrap resamples). Results: Mean age was 38.7 ± 9.8 years, and 60.6% of participants were female. Literacy was moderate (11.2 ± 4.1/20) and familiarity favored GDPR (69.7%) over the EU AI Act (25.7%). Literacy correlated with confidence (=0.52), whereas confidence correlated with adoption intention (=0.41); trust correlated positively (=0.44) and burden correlated negatively (=−0.29) with adoption. High adoption intention was noted in 50.5% of participants and was independently associated with higher literacy (aOR 1.85 per +1 SD; 95% CI 1.20–2.85), higher trust (aOR 1.72; 1.13–2.63), lower burden (aOR 0.64; 0.43–0.95), and prior AI training (aOR 2.10; 1.03–4.29). Three phenotypes emerged (Confident Adopters n = 44; Cautious Compliers n = 36; Skeptical Low Literacy n = 29), with adoption scores of 4.2 ± 0.5 vs. 3.1 ± 0.7 in the highest and lowest groups. Mediation showed a partial indirect effect via confidence (0.13; 95% CI 0.05–0.24). Conclusions: AI legislation literacy, confidence, trust, and perceived burden are key, modifiable determinants of AI adoption intentions; phenotype-guided strategies can target training, governance support, and post-deployment monitoring readiness. The revised framing explicitly situates these determinants within recent AI-specific regulatory and technical developments, including high-risk AI obligations, AI-enabled medical device change control, generative/large multimodal model risks, and lifecycle monitoring. Full article

A dual-platform mushroom cultivation system integrating artificial intelligence (AI)-assisted environmental monitoring and controlled-environment agriculture (CEA) was developed to support experiential STEM education across K–12 and undergraduate settings. Hands-on instruction with multicellular fungi is often limited by reliance on microbial models and by constraints associated with field-based activities. To address this gap, we implemented an indoor instructional platform that combines a commercial AI-assisted automated cultivation unit with a tent-based chamber for hands-on environmental control. Representative cultivated species included oyster mushrooms (Pleurotus spp.) and lion’s mane (Hericium erinaceus). The AI-assisted system provided sensor/camera-based monitoring, app-based feedback, and software-assisted regulation of humidity, light, and airflow, whereas the tent-based system enabled direct student manipulation of cultivation conditions. Together, the systems allowed students to observe fungal development, manage environmental parameters, and collect quantitative and qualitative data within a single academic term. Post-harvest activities, including mushroom-based food preparation and tasting, further connected fungal biology with food and sustainability. A matched pre- and post-course survey (n = 30) showed increases in students’ self-reported perceived understanding, cultivation confidence, and engagement, with mean scores increasing from approximately 2–4 to 6–8. Because the survey instrument was not formally validated and no control group was included, these results are interpreted as preliminary self-reported perceptions rather than objective evidence of learning gains. The platform provides a practical model for integrating fungal biology, AI-assisted environmental monitoring, and CEA into STEM education. Full article

This review traces the four-generation evolution of non-destructive evaluation (NDE 1.0–4.0) and audits where the field genuinely stands today. The central finding is that statistically qualified probability of detection (POD), as defined in MIL-HDBK-1823A and related frameworks, is not interchangeable with machine-learning metrics such as accuracy or F1-score; the two answer different questions and rest on different statistical foundations. Reported AI performance on curated datasets does not, by itself, predict field reliability because domain shift, sensor variability, and class imbalance change the inspection signal once a model leaves the lab. Six recurring barriers limit industrial uptake: scarce open benchmark datasets, domain shift, weak interoperability, explainability constraints, cybersecurity exposure, and the lack of broadly accepted code provisions for AI-derived accept/reject decisions. The oil and gas sector is used as a case study because it combines high inspection volume, severe operating environments, mature risk-based inspection practice, and strong regulatory conservatism. NDE 4.0 is technically credible; its wider acceptance in safety-critical industries will be earned through representative field validation, auditable model governance, standardised data structures, and qualification pathways—not through stronger laboratory accuracy claims. Full article

Renewable integration is increasingly important for sustainable off-grid computing. The inherent variability of solar output frequently produces unusable midday surpluses. Leveraging recent Artificial Intelligence (AI) advances and established literature, we evaluate an AI-driven demand-response framework for scaling Large Language Models (LLMs) training servers using real-time solar energy data, Solcast forecasts, and battery storage records collected from Battery Management Systems (BMS), Maximum Power Point Tracking (MPPT) units, and smart inverters. An n8n AI Agent using the Ollama chat model gpt-oss:20b assesses surplus solar energy, activating selected servers to utilize otherwise wasted capacity. Workloads consistently align with solar availability, demonstrating 99% operational reliability, sub-second responsiveness, and accurate surplus-energy detection. This research demonstrates how Artificial Intelligence can repurpose surplus solar output into usable computational capacity, thereby contributing to a broader transition toward renewable-powered infrastructures. Full article

Conventional chemical analysis relies on costly laboratory instrumentation, while current e-nose systems are expensive for widespread deployment. New opportunities for low-cost, accessible e-nose applications are emerging for diverse fields due to the rapid evolution of inexpensive sensor technologies. We developed a framework that enables rapid integration of newly available low-cost gas sensors into functional e-nose systems, continuously evaluating them as they become commercially available. By characterizing their performance in multi-sensor arrays that mimic biological olfaction, the framework demonstrates effective odor discrimination in a low-cost e-nose system through coordinated behavior of a heterogeneous sensor array. Our testing approach includes sensor sensitivity, selectivity, and stability, which are to be combined with appropriate pattern recognition and AI algorithms in the future for effective chemical discrimination. This work provides a pathway for continuously updating e-nose technology with the latest available sensors in a cost-effective manner, thereby making advanced chemical sensing accessible for resource-limited settings and enabling large-scale deployment in real-world applications with future potential applications such as food quality monitoring, environmental sensing, smart agriculture, etc. Full article

Soil quality is central to agricultural sustainability and food security, yet coastal agroecosystems are increasingly threatened by degradation from intensive practices and seawater intrusion. This study aimed to integrate soil quality index (SQI), statistical modeling, machine learning (ML), and decision analysis to assess and manage soil health in the Skhirat coastal plain of Morocco. A total of 30 topsoil samples were collected and analyzed for chemical and nutrient properties. Spatial interpolation revealed strong coast–inland gradients where EC ranged from 0.47 to 6.3 dS/m with the highest salinity in the south-western fringe, while CEC (8.4–39.7 cmol/kg) and OM (0.54–2.81%) peaked inland. Principal component analysis (PCA) explained 65.9% of total variance, with salinity drivers loading negatively against fertility indicators. Redundancy analysis (RDA) biplots highlighted antagonism between salinity and fertility axes. The PCA-minimum data set (MDS)-SQI integrated key indicators and ranged from 0.084 to 0.897 (mean 0.614), classifying 33% of sites as low quality. The ML model linear regression achieved the best performance (R² = 0.907). Multi-criteria decision analysis (MCDA) using TOPSIS and PROMETHEE II prioritized coastal sites with indices up to 0.882, and robust underweight sensitivity (Spearman ρ = 0.992). This integrated framework demonstrates that soil chemical monitoring, AI prediction, and MCDA can jointly deliver robust, site-specific management strategies for vulnerable coastal agroecosystems. Full article

Photosynthetic biogas upgrading (PBU) using microalgae is a promising biological approach for converting raw biogas into biomethane while recovering nutrients and fixing part of the biogenic CO₂ into algal biomass. Unlike conventional physicochemical technologies, which mainly separate CO₂ from CH₄, PBU can combine gas upgrading with wastewater or digestate treatment, nutrient recycling, and biomass production. This review assesses the current state of PBU technology, with particular emphasis on high-rate algal ponds, absorption columns, and closed photobioreactors. It examines the main operating parameters that control gas–liquid mass transfer, carbonate buffering, and photosynthetic activity, including the liquid-to-gas ratio, pH, alkalinity, temperature, light regime, light intensity, and gas retention time. Special attention is given to the combined effects of the L/G ratio, pH, and alkalinity, as these parameters strongly influence CO₂ absorption, CH₄ enrichment, and O₂ contamination of the upgraded gas. The use of wastewater or anaerobic digestate instead of synthetic growth media is identified as an important sustainability advantage, particularly at wastewater treatment plants with existing anaerobic digestion and nutrient-rich side streams. However, digestate use may also create operational challenges related to turbidity, ammonium inhibition, solids, and variable composition. Available studies indicate that PBU may reduce operating costs and greenhouse gas emissions under favorable conditions while creating additional value from algal biomass. Nevertheless, wider deployment is still limited by high land requirements, seasonal variability, O₂ contamination, biomass harvesting, and limited evidence from large-scale systems. Future development should therefore focus on improved oxygen management, more efficient reactor designs, nanoparticle-assisted enhancement of photosynthetic activity, better integration with wastewater treatment, and AI-supported monitoring and control to improve process stability and support scale-up. Full article

The rapid advancement of machine learning and artificial intelligence (AI) has created new opportunities to enhance diagnostic accuracy in dermatology, particularly within primary care settings. Computer-aided diagnosis (CAD) systems have demonstrated potential to support General Practitioners (GPs) by enabling earlier and more consistent identification of skin diseases. This review critically examines the literature on explainable artificial intelligence (XAI) for skin disease classification, with a specific focus on the evolution of explainability frameworks and the methodological implications of dataset selection. A comprehensive review of studies published between 2020 and 2025 was conducted across multiple academic databases, encompassing research on skin lesion detection, classification, and monitoring. The analysis reveals that deep learning architectures, particularly those leveraging transfer learning with models such as EfficientNet, ResNet, and Xception, frequently report high classification accuracies—often exceeding 90% when evaluated on single benchmark datasets. However, studies employing multiple datasets consistently demonstrate more stable and generalisable performance, albeit with modest reductions in reported accuracy, highlighting a critical trade-off between performance optimisation and real-world robustness. The review further identifies a clear temporal progression in the adoption of XAI techniques. Early studies relied on a broader range of post hoc explainability while later work increasingly consolidated around Grad-CAM, SHAP, and related attribution techniques, followed by gradual diversification into more specialised frameworks such as TCAVs (Testing with Concept Activation Vectors) and Prototype-based Networks. Despite these advances, the lack of clinically grounded explanations, limited integration of ethical considerations, and reliance on non-clinical imagery continue to constrain clinical applicability which we have explored using a GRADE-style narrative. Notably, evidence suggests that CAD systems can improve GP diagnostic accuracy for conditions such as melanoma and seborrhoeic keratosis; however, sustained clinical adoption remains contingent on transparent, reliable, and context-aware explainability mechanisms. Full article

Offshore wind farms are increasingly and rapidly expanding due to their ability to harness strong and consistent wind energy resources. Large offshore wind farms are connected to mainland grids through High-Voltage Direct Current (HVDC) technology. However, offshore wind farms can often experience disturbances related to sudden wind changes, voltage drops/dips, faults related to converter switching, and unbalanced grid conditions which affect both the HVDC operation and wind turbine output. As a result, there is a growing need for more advanced and reliable modeling and monitoring tools. Moreover, traditional proportional-integral (PI) controllers are widely applied in wind turbines and HVDC systems due to their simple structure, easy implementation, and reliability. However, PI controllers perform poorly under non-linear and abnormal/fast-changing conditions, especially during sudden drops in wind power and grid faults. With this background, this paper first develops a digital twin model of an offshore wind farm that enables remote operation and monitoring of individual wind turbines. Also, an artificial intelligence (AI)-based controller, namely a fuzzy logic controller (FLC), is proposed to maintain transient stability of a full digital twin-based offshore wind farm connected to the HVDC grid under fault conditions. The effectiveness of the proposed FLC is demonstrated by considering a digital twin-enabled 700 MW offshore wind farm. The performance of the proposed FLC has been compared with that of the PI controller. Simulations performed by the MATLAB/Simulink software show that during the moderate voltage dip at 15 s, the PI controller experienced a 29.8% power reduction with a recovery time of approximately 9 s, whereas the FLC reduced the power drop to 23.1% and recovered within 6 s. During the severe converter disturbance at 15 s, the PI controller recorded a 36.9% power reduction compared to 23.4% for the FLC. Similarly, during the short-duration turbulence at 15 s, the PI controller exhibited a 36.73% power drop and recovered in approximately 7 s, while the FLC limited the power reduction to 19.17% and recovered within 5s. Overall, the FLC provided improved voltage stability, faster recovery, reduced oscillations, and superior fault ride-through capability compared with the conventional PI controller, demonstrating its effectiveness for digital twin-enabled offshore wind farm application. Full article