Search Results (12,457)

Grasslands represent one of the most extensive terrestrial biomes globally, covering approximately one-third of the Earth’s land surface, yet they are increasingly threatened by land-use change and overgrazing, underscoring the need for reliable monitoring approaches. This study compares the performance of machine learning and deep learning algorithms for grassland mapping using multi-source remote sensing data derived from Sentinel-1, Sentinel-2, and terrain variables. The research was conducted in Mpumalanga Province, South Africa, a heterogeneous landscape comprising lowland savannas, high-altitude grasslands, escarpments, and riverine wetlands. Random Forest (RF) and Support Vector Machine (SVM) classifiers were implemented in Google Earth Engine using fused satellite and terrain datasets with field-collected samples for training and validation, while a One-Dimensional Convolutional Neural Network (1D-CNN) was developed in Python 3.13.5 using the same inputs. Results demonstrate that integrating multi-source data improves classification accuracy, with radar-based features contributing the most. RF achieved the highest performance, with an overall accuracy of 97.7% and grass-class precision, recall, and F1-score exceeding 0.97, closely followed by the 1D-CNN with 91% overall accuracy and complete grass detection. In contrast, SVM performed notably lower with an overall accuracy of 80,8%. These findings highlight the effectiveness of advanced learning approaches for grassland mapping and support their application in ecological restoration and environmental management. Full article

The production of “winter wines” in south-eastern Brazil, enabled by the double pruning technique, requires precise assessment of grape technological maturity to ensure high-quality outputs. However, conventional monitoring approaches are destructive, labor-intensive, and limited in their ability to capture vineyard spatial variability. This study aimed to develop and validate a non-destructive predictive framework for Soluble Solids (°Brix) and Titratable Acidity (TA) by integrating spatial remote sensing data with temporal agrometeorological information. Multispectral imagery was acquired via an unmanned aerial vehicle in a vineyard cultivated with Sauvignon Blanc and Syrah, from which vegetation indices were derived and combined with Growing Degree-Days to train machine learning models, including Random Forest, Multilayer Perceptron, and XGBoost. The incorporation of agrometeorological data significantly improved predictive performance compared to models based solely on vegetation indices. Among the tested algorithms, XGBoost achieved the highest accuracy, with coefficients of determination of 0.89 for °Brix and 0.77 for TA, achieved by XGBoost on an independent hold-out test set. Model interpretability analysis indicated that Growing Degree-Days and cultivar were the primary drivers of maturation dynamics, while vegetation indices refined predictions by accounting for spatial variability in plant vigor. Overall, the proposed approach represents a promising proof-of-concept framework for non-destructive maturity monitoring in precision viticulture, supporting improved monitoring of grape maturation. However, multi-season validation across diverse vineyard conditions is required to confirm its generalizability and support its application as a routine decision-support tool. Full article

Conventional object detection models suffer from significant performance degradation in dense urban traffic scenarios. To address these critical limitations and enable accurate real-time road congestion decision making, this study proposes an optimized YOLOv8-based detection paradigm that decouples multi-scale feature enhancement from dynamic focused bounding box regression. Specifically, a multi-scale feature enhancement (MFE) module is designed to extract high-resolution shallow features directly from the P2 layer of the YOLOv8 backbone. Then, a convolutional block attention module (CBAM) is embedded into the feature fusion neck to adaptively filter complex urban background noise and recalibrate channel–spatial feature responses for vehicle target saliency. Furthermore, the standard CIoU loss is replaced with the Wise-IoU (WIoU) dynamic focusing loss function, which suppresses gradient interference from low-quality, occluded samples and stabilizes bounding box regression for dense vehicle targets. The high-precision vehicle detection outputs are fed into a quantitative congestion index (CI) model, which fuses vehicle density and average speed to realize real-time congestion-level classification. Extensive experiments on the public UAVDT benchmark dataset demonstrate that the proposed model achieves an mAP@0.5 of 83.1% (3.8 percentage points higher than the YOLOv8 baseline), an mAP_S (small target) of 23.2% (a 4.3 percentage point improvement), and a real-time congestion decision accuracy of 83.8%. Ablation studies verify the independent and synergistic effectiveness of the MFE, CBAM, and WIoU modules, with the MFE module making the greatest contribution to small-target detection performance (+1.7% mAP@0.5). The proposed model maintains a real-time inference speed of 86 FPS (frames per second) on an NVIDIA RTX 3090 GPU, far exceeding the 30 FPS threshold for real-time traffic monitoring. Full article

Freshwater is essential for sustaining the health of humans, animals, and ecosystems; however, agricultural activities remain a major source of water pollution globally. This review examines how crop production, livestock farming, and aquaculture contribute to water contamination, the effectiveness of current European policies, and the potential of sustainable mitigation strategies. Evidence from the research identified pesticides, herbicides, veterinary antibiotics, nutrient runoff, aquaculture effluents, and microplastics as the primary agricultural pollutants affecting surface and groundwater quality. These contaminants have been linked to ecosystem degradation, biodiversity loss, endocrine disruption, antimicrobial resistance, and adverse human health outcomes. Despite extensive regulatory frameworks, including the Water Framework Directive, Nitrates Directive, Farm to Fork Strategy, and European Green Deal, significant implementation and monitoring challenges remain. Current evidence indicates that only 40% of European surface waters achieve “good” ecological status, highlighting persistent water quality concerns across the region. The review further identified precision irrigation, Internet of Things (IoT)-enabled monitoring, biopesticides, hydroponic systems, and integrated multi-trophic aquaculture as promising solutions for reducing agricultural impacts on water resources. However, barriers, including high implementation costs, technological limitations, and inconsistent policy enforcement, continue to hinder widespread adoption. Overall, the findings demonstrate that while existing policies have improved water governance, stronger regulatory enforcement, greater investment in sustainable technologies, and increased adoption of nature-based solutions are required to reduce agricultural water pollution. An integrated approach combining technological innovation, policy support, and sustainable farming practices is essential to protect freshwater resources and ensure long-term environmental sustainability. Full article

Cloud-dependent irrigation platforms can support remote monitoring, but their use in precision fertigation is limited when local decisions must be made quickly and reliably. Network delay, temporary disconnection, and the use of single-point measurements may all reduce the ability of a system to respond to spatial variation in soil moisture and nutrient demand. In this work, an edge-controlled precision fertigation system was developed by combining multi-parameter soil sensing, spatial task scheduling, and a 6-DOF robotic manipulator. The ESP32 controller runs a preemptive FreeRTOS scheduler, allowing sensor acquisition, inverse-kinematics calculation, and pump actuation to be handled as separate tasks. A Kalman filter was used to smooth soil moisture measurements, and a hysteresis-based control strategy was adopted to reduce false triggering and repeated pump switching. To improve fertigation safety, a hardware–software interlock was added so that fertilizer delivery is always accompanied by water delivery. Hardware-in-the-Loop simulation and a 14-day field deployment were used to evaluate the system. The controller achieved an end-to-end latency of less than 38 ms and maintained operation during network interruptions through cached local parameters. After calibration, the robotic end-effector positioning error was reduced to ±2.4 mm. The hysteresis strategy lowered daily pump cycling by 71%. Based on prototype duty-cycle data and seasonal extrapolation, the projected seasonal water use and fertilizer demand were 44% and 38% lower, respectively, than those estimated for a uniform application. These values should be interpreted as model-based projections rather than direct season-long measurements. During 72 h of continuous operation, no Modbus faults were observed, and RTOS heap fragmentation remained stable. Overall, the results suggest that edge-based deterministic control can provide a practical route for precision fertigation where both spatial variability and intermittent connectivity must be considered. Full article

Underwater optical sensor networks are essential for fish monitoring, yet imagery is often affected by illumination variability, low contrast, and complex backgrounds. Attention mechanisms are vital for feature representation in deep networks, yet existing approaches often struggle with spatial information loss and limited multi-scale interaction under such challenging conditions. This paper introduces Convolution to Interactive Capture and Recalibration Enhancement (C2ICARE), a lightweight attention module designed to overcome these challenges. The principal contribution of C2ICARE is the adaptation of memory interaction principles into an edge-oriented attention framework that enhances feature discrimination while maintaining computational efficiency. The architecture employs three core innovations: a 1:3 memory-feature split to preserve context while reducing cost, parallel multi-scale depthwise convolutions (3 × 3 and 7 × 7) for fine-grained and broad feature extraction, and a cross-branch interaction mechanism coupled with a ConvNeXt-style feed-forward network that avoids dimensionality reduction. Experimental results on an underwater fish dataset demonstrate that YOLO26n with C2ICARE achieves a mean average precision (mAP@0.5:0.95) of 0.7033, outperforming Coordinate Attention (+3.8%), FasterBlock (+1.7%), and CBAM (+0.4%) while adding only 0.05M parameters and 0.16 GFLOPs. Multi-objective Pareto Frontier analysis confirms that C2ICARE provides an effective balance between accuracy, efficiency, and generalization for resource-constrained deployment. EigenCAM visualizations further validate that the model focuses on biological morphology rather than background noise. Its lightweight design enables seamless integration with underwater sensor networks and fog platforms for real-time fish detection in aquaculture, commercial fisheries, and scientific research. Future work will investigate broader marine applications and cross-platform deployment scenarios. The code is available on GitHub. Full article

Accurate change detection in high-resolution remote sensing imagery is essential for urban planning, land-use monitoring, and disaster response. This study introduces CTA-Net, a Cross-Temporal Attention Network for binary change detection in bi-temporal optical imagery, designed to improve robustness against pseudo-changes caused by illumination variation, seasonal effects, and sensor noise. The proposed method employs a shared Siamese encoder with multi-scale Cross-Temporal Attention modules that derive spatial and channel attention from L2 feature differences, along with a lightweight confidence estimation head for per-pixel uncertainty modelling. A hybrid loss function combining confidence-weighted binary cross-entropy and focal loss is used to address class imbalance. Experiments on the LEVIR-CD dataset demonstrate that CTA-Net achieves an overall accuracy of 98.99%, an F1-score of 87.68%, an Intersection over Union of 78.06%, a Cohen’s kappa of 0.8715, and a Matthews Correlation Coefficient of 0.8721, with stable convergence and minimal overfitting. Qualitative and calibration analyses further indicate that the model produces interpretable attention maps and reliable probabilistic outputs. To evaluate cross-domain generalization, we conduct a transfer learning case study on multispectral Sentinel-2 agricultural imagery. The model is adapted to 11-channel input and fine-tuned on automatically generated change masks derived from NDVI-delta thresholding. Under this supervision protocol, CTA-Net achieves an F1-score of 95.18% and an IoU of 90.81% on a held-out test region, with balanced precision and recall. While these results demonstrate effective adaptation across sensor modality, spatial resolution, and semantic domain, the evaluation reflects agreement with the mask generation procedure rather than independently annotated ground truth. While CTA-Net shows strong performance and reasonable interpretability, its cross-domain evaluation is limited by the use of automatically generated labels. As a result, the reported transferability should be interpreted cautiously until validated on human-annotated datasets. Full article

As climate change intensifies heat extremes, the Urban Heat Island (UHI) effect amplifies local thermal stress. Assessing the UHI using robust observational data, whether ground- and/or satellite-based, is essential for climate risk assessment and evidence-based urban adaptation. Therefore, this study aims to provide a comprehensive climatological assessment of air temperature patterns and UHI intensity across the Lisbon Metropolitan Area (LMA) over a 26-year period (2000–2025). The methodology employs a dense, high-quality integrated network of in-situ weather stations from the Portuguese Institute for Sea and Atmosphere (IPMA) and the National Water Resources Information System (SNIRH). To bridge critical gaps in traditional climate assessments, this research implements a dual-perspective approach that combines the high temporal resolution of MSG-SEVIRI and the spatial precision of MODIS Land Surface Temperature (LST). This framework accurately captures the lag effects between surface heating and atmospheric response. Validation results demonstrate that satellite-derived LST is a robust proxy for monitoring the nocturnal UHI, with differences generally below 1 °C compared with near-surface air temperature observations (T2m). However, daytime LST significantly overestimates atmospheric temperatures, with deviations of 2–8 °C due to solar radiation and urban geometry. The selection of rural reference stations constitutes a critical methodological factor, as a baseline shift can alter perceived UHI intensities by more than 3 °C. Despite these sensitivities, the results unequivocally confirm a persistent and spatially heterogeneous UHI effect in Lisbon, which intensifies during extreme heat events by up to an additional 4 °C. Analysis of the 2003 and 2018 heatwaves reveals surface LST anomalies exceeding 10 °C and urban–rural thermal differentials reaching up to 7 °C under conditions of suppressed maritime breezes. These nocturnal anomalies are particularly pronounced in densely built-up areas, limiting thermal dissipation and preventing physiological recovery. Integrating multi-sensor satellite data with in-situ validation provides a new benchmark for climate risk assessments, delivering the reliable, reproducible data required to strengthen long-term urban resilience under increasingly frequent extreme heat events. Full article

Accurate dairy cow behavior recognition is important for health monitoring, welfare assessment, and early warning in smart livestock farming. However, recognizing fine-grained behaviors such as feeding, drinking, and rumination remains difficult in real barns because of occlusion, complex backgrounds, subtle motion changes, and class imbalance. This study proposes a behavior recognition method that integrates temporal modeling and behavioral priors. The Contrastive Language–Image Pre-training (CLIP) visual encoder is used as the feature extraction backbone, while two temporal adapters are introduced to model dynamic information across consecutive video frames. Dairy cow behavior recognition is further decoupled into posture recognition and action recognition, and a behavioral prior loss is designed to softly constrain unlikely posture–action combinations, such as lying with feeding or lying with drinking. On the test set, the proposed method achieves a five-class accuracy of 75.45%, a five-class Macro-F1 of 0.7246, and an Action Macro-F1 of 0.7605, outperforming the CLIP baseline and several representative video recognition models. These results indicate that the proposed method can support non-contact monitoring of key dairy cow behaviors for practical barn management. Full article

This study presents a comprehensive analysis of the modeling and optimization of sensor installation nodes for weight measurement in the hoppers of a charging car utilized in coke production. The research highlights the critical role of precise load monitoring in preventing technological disruptions, minimizing equipment degradation, and optimizing energy consumption. Conventional sensor technologies, including capacitive, ultrasonic, and laser-based systems, are evaluated, with weight sensors mounted on hopper supports identified as the most robust solution for real-time mass determination under industrial conditions characterized by high dust levels, temperature fluctuations, and mechanical vibrations. A finite element analysis (FEA) was conducted to assess the structural behavior of sensor installation nodes under three distinct loading scenarios, corresponding to different operational conditions of the charging car. The four-point support structure of the hopper experienced the highest loads and non-uniformities. A stress–strain analysis of the sensor mounting assembly, performed using the Ansys software package, confirmed that both the sensor and its support structure maintain a sufficient safety margin (version 2024 R1, Ansys Inc., Canonsburg, PA, USA, the academic license provided to Wrocław University of Science and Technology). The findings validate the structural integrity and operational reliability of the proposed sensor configuration, contributing to the advancement of automated monitoring and control systems in coke production. Full article

Severe air leakage in the goaf of gob-side entry retaining panels can intensify oxygen supply to residual coal and consequently increase the probability of coal spontaneous combustion. Taking the 3451S working face of a coal mine in Hebei Province as the engineering case, this study integrated in situ beam-tube monitoring with Fluent-based numerical simulation to characterize the evolution of the spontaneous-combustion three zones and to optimize prevention and control measures. The results demonstrate that the oxidation zone is characterized by an inclined, continuous band-like distribution penetrating the goaf. The simulated oxygen distribution is consistent with the field measurements, demonstrating the reliability of the established numerical model. The ventilation pattern markedly affects the air-leakage flow field and oxygen concentration distribution, and the Y-type ventilation mode exhibits a higher spontaneous-combustion risk. When the air-volume ratio between the 3451S haulage roadway and the gob-side retained entry is adjusted to 3:1, the oxidation-zone area decreases by approximately 11%. A combined control strategy involving cement-blanket and polymer-spraying leakage sealing, together with precise nitrogen injection, is then proposed to improve the goaf oxygen environment. At a nitrogen-injection rate of 600 m³/h, the oxidation-zone area is reduced by 11,160 m² and the CO concentration remains stable at approximately 4.9 ppm, providing field evidence for improved fire-prevention performance. These results support the design of targeted spontaneous-combustion control strategies for gob-side entry retaining goafs. Full article

Using accurate land cover data is essential to monitor land use and cover changes and assess the effectiveness of various environmental policies. This study evaluates the accuracy of contemporary global land cover products with 10 m spatial resolution, including Google’s Dynamic World (GDW), European Space Agency’s World Cover (ESA WC) and Esri Land Cover (ELC) in mapping forested areas in Poland, aiming to test an assumption if the combination of these products may improve forest mapping accuracy compared to any individual product. Three global datasets and their combinations were assessed with the 2022 EU Land Use/Cover Area Frame Survey (LUCAS). A land cover map of Poland (S2GLC PL) for 2021 served as an auxiliary reference data set. Forest cover classification accuracy was evaluated using precision, recall, and F1-score metrics, and spatial agreement of binary forest maps in the thematic global products was measured with the Intersection over Union (IoU) at two various scale levels (country and province). Our results showed that forest mapping accuracy of three global products varies for Poland, with F1-score equal to 72.2% for ELC, 76.9% for ESA WC, and 68.8% for GDW. IoU against S2GLC PL was equal to 82.6%, 82.3% and 75.2%, for ELC, ESA WC and GDW, respectively, and slightly exceeded 70.5% for three global products. A specific combination of binary forest maps from global products, where the output forest area consisted of forests mapped at the same time by all three products and forests mapped at the same time only by GDW and ESA WC yielded better accuracy indicators than any single product and other tested combinations (F1-score equal to 80.4%, and IoU against S2GLC PL equal to 87.1%). Full article

In intensive aquaculture, non-invasive real-time monitoring of morphological traits and body weight of largemouth bass (Micropterus salmoides) is essential for precision feeding and yield estimation. Manual measurement is laborious and stressful, whereas vision-based methods are challenged by insufficient dual-view feature fusion, poor robustness to underwater noise, and over-parameterized models unsuitable for edge deployment. To address these issues, a lightweight framework, MECT-MobileViT, is proposed based on MobileViT-xxs. A Morphometric-Guided Multi-Scale Fusion module is designed to couple physical priors with dual-branch visual features, strengthening shape–weight association. An ECA-NL attention block employing instance normalization, GLU gating, and threshold filtering is embedded to enhance feature robustness against visual disturbances typical in aquaculture and to accentuate critical morphological features. A three-stage synergistic pruning strategy—attention head pruning, structured channel pruning, and depthwise separable attention substitution—is applied to achieve substantial compression while preserving representational capacity. Experiments on a self-built lateral–dorsal dual-view dataset show that the proposed model significantly outperforms mainstream benchmarks. The pruned version attains an R² of 0.8266 and an RMSE of 16.4201, with less than 2% accuracy degradation relative to the best unpruned model, and contains only 7.34 M parameters. This study demonstrates a promising prototype for contactless, stress-free weight estimation in largemouth bass and offers new technical insights into feature fusion, noise suppression, and collaborative model compression for aquaculture visual perception. Full article

Idelalisib (IDE) is a marketed chiral anticancer drug administered as the S-enantiomer, requiring sensitive monitoring of the R-enantiomer to ensure enantiomeric purity. However, no dedicated capillary electrophoresis (CE) method has been reported for trace-level quantification of R-IDE. In this study, a cyclodextrin-mediated CE method was developed for reliable detection of the R-enantiomer at the 0.1% level (LOD 2 µg/mL; LOQ 5 µg/mL). Systematic screening identified hydroxypropyl-β-cyclodextrin (HP-β-CD) with an intermediate degree of substitution (DS~6.8) as the optimal chiral selector, providing efficient enantioseparation (R_s up to 4.3). The method was validated according to ICH Q2(R2) guidelines, demonstrating suitable precision, accuracy, and robustness. Complementary NMR studies revealed hindered rotation of the 3-phenyl moiety and elucidated the molecular basis of enantioselectivity. Complexation with β-CD and HP-β-CD produced clear diastereomeric differentiation in both ¹H and ¹⁹F NMR spectra, while the simplified ¹⁹F NMR profiles enabled direct enantiomer discrimination. NOESY and ROESY experiments demonstrated distinct inclusion modes, with HP-β-CD accommodating both the fluorinated aromatic ring and the 3-phenyl moiety. These interactions may account for the superior enantioseparation observed with HP-β-CD of intermediate DS. Our validated CE method addresses the distomer determination while NMR insights provide mechanistic understanding of the chiral recognition. Full article

Pig vocalization recognition can support non-invasive monitoring in precision livestock farming, but labelled pig-sound recordings are often limited for specific behaviours or physiological states. Under few-shot conditions, deep models may overfit, whereas traditional acoustic features may not fully describe class-specific time-frequency patterns. This study proposed PSA-AP, a pig-sound adaptation pipeline that uses log-Mel spectrograms and integrates SpecAugment-based domain expansion, ImageNet-pretrained ResNet18 knowledge transfer, and ArcFace-based feature alignment. The method was designed to reduce dependence on limited labelled samples, improve task-adapted representation learning, and enhance inter-class separability in the embedding space. Experiments were conducted on a five-class few-shot pig vocalization classification task, including eat, estrous, farrowing (fap), howl, and oink sounds collected from 10 adult Landrace pigs. Using $K=\{5,10,15,20,25,30\}$ labelled wav files per class and five random seeds, each selected training wav file and each held-out test wav file was converted into one 1.0 s log-Mel spectrogram for model training or evaluation. Final evaluation was based on the last checkpoint of each training run. PSA-AP achieved the best mean Accuracy, Macro-F1, and UAR at every K-shot setting. At $K=30$, PSA-AP reached 90.60% Accuracy, 90.49% Macro-F1, and 90.60% UAR, exceeding Raw by 7.80, 7.82, and 7.80 percentage points, respectively. These results indicate that the proposed integration of domain expansion, knowledge transfer, and feature alignment provides a feasible supervised adaptation strategy for few-shot pig vocalization recognition within the current protocol. Full article