Search Results (2,661)

Crop disease identification is a pivotal research area in smart agriculture, forming the foundation for disease mapping and targeted prevention strategies. Among the most prevalent global wheat diseases, powdery mildew—caused by fungal infection—poses a significant threat to crop yield and quality, making early and accurate detection crucial for effective management. In this study, we present QY-SE-MResNet34, a deep learning-based classification model that builds upon ResNet34 to perform multi-class classification of wheat leaf images and assess powdery mildew severity at the single-leaf level. The proposed methodology begins with dataset construction following the GBT 17980.22-2000 national standard for powdery mildew severity grading, resulting in a curated collection of 4248 wheat leaf images at the grain-filling stage across six severity levels. To enhance model performance, we integrated transfer learning with ResNet34, leveraging pretrained weights to improve feature extraction and accelerate convergence. Further refinements included embedding a Squeeze-and-Excitation (SE) block to strengthen feature representation while maintaining computational efficiency. The model architecture was also optimized by modifying the first convolutional layer (conv1)—replacing the original 7 × 7 kernel with a 3 × 3 kernel, adjusting the stride to 1, and setting padding to 1—to better capture fine-grained leaf textures and edge features. Subsequently, the optimal training strategy was determined through hyperparameter tuning experiments, and GrabCut-based background processing along with data augmentation were introduced to enhance model robustness. In addition, interpretability techniques such as channel masking and Grad-CAM were employed to visualize the model’s decision-making process. Experimental validation demonstrated that QY-SE-MResNet34 achieved an 89% classification accuracy, outperforming established models such as ResNet50, VGG16, and MobileNetV2 and surpassing the original ResNet34 by 11%. This study delivers a high-performance solution for single-leaf wheat powdery mildew severity assessment, offering practical value for intelligent disease monitoring and early warning systems in precision agriculture. Full article

Background/Objectives: To assess the association between early pregnancy carbohydrate quality, as measured by the Carbohydrate Quality Index (CQI), and the risk of delivering a large-for-gestational-age (LGA) infant in a Mediterranean pregnant cohort of northern Greece. Methods: We analyzed singleton pregnancies from the BORN 2020 prospective cohort in Greece. Dietary intake was assessed via a validated food frequency questionnaire, and CQI was computed from glycemic index, fiber density, whole-to-refined grain ratio, and solid-to-liquid carbohydrate ratio. Multivariable logistic regression was used to estimate the association between CQI (in tertiles) and LGA risk, defined as birthweight >90th percentile. Results: Among the 797 participants, 152 (19.1%) delivered LGA infants, and 117 (14.7%) were diagnosed with GDM. Of those with GDM, 23 (19.7%) delivered LGA infants. In the total population, higher maternal weight (p < 0.001), height (p = 0.006), and pre-pregnancy BMI (p = 0.004) were significantly associated with LGA. A greater proportion of women with LGA had a BMI > 25 (p = 0.007). In the GDM subgroup, maternal height remained significantly higher in those who delivered LGA infants (p = 0.017). In multivariable models, moderate CQI was consistently associated with increased odds of LGA across all models (Model 1: aOR = 1.60 (95% CI: 1.03–2.50), p = 0.037, Model 2: aOR = 1.57 (95% CI: 1.01–2.46), p = 0.046, Model 3: aOR = 1.58 (95% CI: 1.01–2.47), p = 0.044, Model 4 aOR: 1.70; 95% CI: 1.08–2.72; p = 0.023), whereas high CQI was not. In the GDM subgroup, a significant association between high CQI and increased LGA risk was observed in less adjusted models (Model 1 aOR: 6.74; 95% CI: 1.32–56.66; p = 0.039, Model 2 aOR: 6.64; 95% CI: 1.27–57.48; p = 0.044), but this was attenuated and became non-significant in the fully adjusted model (aOR: 3.05; 95% CI: 0.47–30.22; p = 0.28). When examining CQI components individually, no consistent associations were observed. Notably, a higher intake of low-quality carbohydrates (≥50% of energy intake) was significantly associated with increased LGA risk in the total population (aOR: 4.25; 95% CI: 1.53–11.67; p = 0.005). Conclusions: Higher early pregnancy intake of low-quality carbohydrates was associated with an elevated risk of LGA in the general population. However, CQI itself showed a non-linear and inconsistent relationship with LGA, with moderate, but not high, CQI linked to increased risk, particularly in GDM pregnancies, where associations were lost after adjustment. Both carbohydrate quality and quantity evaluations are essential, particularly in high-risk groups, to inform dietary guidance in pregnancy. Full article

Hyperspectral sparse unmixing (SU) networks have recently received considerable attention due to their model hyperspectral images (HSIs) with a priori spectral libraries and to capture nonlinear features through deep networks. This method effectively avoids errors associated with endmember extraction, and enhances the unmixing performance via nonlinear modeling. However, two major challenges remain: the use of large spectral libraries with high coherence leads to computational redundancy and performance degradation; moreover, certain feature extraction models, such as Transformer, while exhibiting strong representational capabilities, suffer from high computational complexity. To address these limitations, this paper proposes a hyperspectral unmixing dual-branch network integrating an adaptive hop-aware GCN and neighborhood offset Mamba that is termed DGMNet. Specifically, DGMNet consists of two parallel branches. The first branch employs the adaptive hop-neighborhood-aware GCN (AHNAGC) module to model global spatial features. The second branch utilizes the neighborhood spatial offset Mamba (NSOM) module to capture fine-grained local spatial structures. Subsequently, the designed Mamba-enhanced dual-stream feature fusion (MEDFF) module fuses the global and local spatial features extracted from the two branches and performs spectral feature learning through a spectral attention mechanism. Moreover, DGMNet innovatively incorporates a spectral-library-pruning mechanism into the SU network and designs a new pruning strategy that accounts for the contribution of small-target endmembers, thereby enabling the dynamic selection of valid endmembers and reducing the computational redundancy. Finally, an improved ESS-Loss is proposed, which combines an enhanced total variation (ETV) with an $l_{1/2}$ sparsity constraint to effectively refine the model performance. The experimental results on two synthetic and five real datasets demonstrate the effectiveness and superiority of the proposed method compared with the state-of-the-art methods. Notably, experiments on the Shahu dataset from the Gaofen-5 satellite further demonstrated DGMNet’s robustness and generalization. Full article

In order to improve the microstructure and corrosion resistance of entropy alloy in the FeCrNi system, laser melting deposition technology was used as a preparation method to study the effects of different contents of Al and Ti on the microstructure and corrosion resistance of entropy alloy in FeCrNi(AlTi)_x (x = 0.17, 0.2, and 0.24). The results show that the addition of Al and Ti elements can change the phase structure of the alloy from a single FCC phase structure to an FCC + BCC biphase structure. The BCC phase volume fraction of FeCrNi(AlTi)_0.2 is the highest among the three alloys, reaching 37.5%. With the addition of Al and Ti content, the grain of the alloy will be refined to a certain extent. In addition, the dual-phase structure will also improve the corrosion resistance of the alloy. In 3.5 wt.% NaCl solution, the increase of Al and Ti content can effectively improve the protection of the passivation film on the surface of the entropy alloy in FeCrNi(AlTi)_x, effectively inhibit the large-scale corrosion phenomenon on the alloy surface, and thus improve the corrosion resistance of the alloy. In a certain range, increasing the content of Al and Ti elements in the FeCrNi(AlTi)_x system can improve the corrosion resistance of the alloy. Full article

Motor Imagery (MI) classification plays a crucial role in enhancing the performance of brain–computer interface (BCI) systems, thereby enabling advanced neurorehabilitation and the development of intuitive brain-controlled technologies. However, MI classification using electroencephalography (EEG) is hindered by spatiotemporal variability and the limited interpretability of deep learning (DL) models. To mitigate these challenges, dropout techniques are employed as regularization strategies. Nevertheless, the removal of critical EEG channels, particularly those from the sensorimotor cortex, can result in substantial spatial information loss, especially under limited training data conditions. This issue, compounded by high EEG variability in subjects with poor performance, hinders generalization and reduces the interpretability and clinical trust in MI-based BCI systems. This study proposes a novel framework integrating channel dropout—a variant of Monte Carlo dropout (MCD)—with class activation maps (CAMs) to enhance robustness and interpretability in MI classification. This integration represents a significant step forward by offering, for the first time, a dedicated solution to concurrently mitigate spatiotemporal uncertainty and provide fine-grained neurophysiologically relevant interpretability in motor imagery classification, particularly demonstrating refined spatial attention in challenging low-performing subjects. We evaluate three DL architectures (ShallowConvNet, EEGNet, TCNet Fusion) on a 52-subject MI-EEG dataset, applying channel dropout to simulate structural variability and LayerCAM to visualize spatiotemporal patterns. Results demonstrate that among the three evaluated deep learning models for MI-EEG classification, TCNet Fusion achieved the highest peak accuracy of 74.4% using 32 EEG channels. At the same time, ShallowConvNet recorded the lowest peak at 72.7%, indicating TCNet Fusion’s robustness in moderate-density montages. Incorporating MCD notably improved model consistency and classification accuracy, especially in low-performing subjects where baseline accuracies were below 70%; EEGNet and TCNet Fusion showed accuracy improvements of up to 10% compared to their non-MCD versions. Furthermore, LayerCAM visualizations enhanced with MCD transformed diffuse spatial activation patterns into more focused and interpretable topographies, aligning more closely with known motor-related brain regions and thereby boosting both interpretability and classification reliability across varying subject performance levels. Our approach offers a unified solution for uncertainty-aware, and interpretable MI classification. Full article

In cybersecurity, constructing an accurate knowledge graph is vital for discovering key entities and relationships in security incidents buried in vast unstructured threat reports. Traditional knowledge-graph construction pipelines based on handcrafted rules or conventional machine learning models falter when the data scale and linguistic variety grow. GraphRAG, a retrieval-augmented generation (RAG) framework that splits documents into fixed-length chunks and then retrieves the most relevant ones for generation, offers a scalable alternative yet still suffers from fragmentation and semantic gaps that erode graph integrity. To resolve these issues, this paper proposes SC-LKM, a cybersecurity knowledge-graph construction method that couples the GraphRAG backbone with hierarchical semantic chunking. SC-LKM applies semantic chunking to build a cybersecurity knowledge graph that avoids the fragmentation and inconsistency seen in prior work. The semantic chunking method first respects the native document hierarchy and then refines boundaries with topic similarity and named-entity continuity, maintaining logical coherence while limiting information loss during the fine-grained processing of unstructured text. SC-LKM further integrates the semantic comprehension capacity of Qwen2.5-14B-Instruct, markedly boosting extraction accuracy and reasoning quality. Experimental results show that SC-LKM surpasses baseline systems in entity-recognition coverage, topology density, and semantic consistency. Full article

Fine-grained visual categorization (FGVC) presents significant challenges due to subtle inter-class variation and significant intra-class diversity, often leading to limited discriminative capacity in global representations. Existing methods inadequately capture localized, class-relevant features across multiple semantic levels, especially under complex spatial configurations. To address these challenges, we introduce a Multi-scale Attention-driven Hierarchical Learning (MAHL) framework that iteratively refines feature representations via scale-adaptive attention mechanisms. Specifically, fully connected (FC) classifiers are applied to spatially pooled feature maps at multiple network stages to capture global semantic context. The learned FC weights are then projected onto the original high-resolution feature maps to compute spatial contribution scores for the predicted class, serving as attention cues. These multi-scale attention maps guide the selection of discriminative regions, which are hierarchically integrated into successive training iterations to reinforce both global and local contextual dependencies. Moreover, we explore a generalized pooling operation that parametrically fuses average and max pooling, enabling richer contextual retention in the encoded features. Comprehensive evaluations on benchmark FGVC datasets demonstrate that MAHL consistently outperforms state-of-the-art methods, validating its efficacy in learning robust, class-discriminative, high-resolution representations through attention-guided hierarchical refinement. Full article

The titanium matrix composites (TMCs) fabricated via Directed Energy Deposition (DED) effectively overcome the issue of coarse columnar grains typically observed in additively manufactured titanium alloys. In this study, systematic annealing heat treatments were applied to in situ (TiB + TiC)/Ti-6Al-4V composites to refine the microstructure and tailor mechanical properties. The results reveal that the plate-like α phase in the as-deposited composites gradually transforms into an equiaxed morphology with increasing annealing temperature and holding time. Notably, when the annealing temperature exceeds 1000 °C, significant coarsening of the TiC phase is observed, while the TiB phase remains morphologically stable. Annealing promotes decomposition of acicular martensite and stress relaxation, leading to a reduction in hardness compared to the as-deposited state. However, the reticulated distribution of the TiB and TiC reinforcement phases contributes to enhanced tensile performance. Specifically, the as-deposited composite achieves a tensile strength of 1109 MPa in the XOY direction, representing a 21.6% improvement over the as-cast counterpart, while maintaining a ductility of 2.47%. These findings demonstrate that post-deposition annealing is an effective strategy to regulate microstructure and achieve a desirable balance between strength and ductility in DED-fabricated titanium matrix composites. Full article

Hydrotechnical structures reshape sandy coasts by altering hydrodynamics and sediment transport, yet their long-term effects on sediment texture remain underexplored, particularly in the Baltic Sea. This study investigates the spatial and temporal variations in sediment grain size near two ports (Šventoji and Klaipėda) on the sandy Baltic Sea coast, considering the influence of jetties, nourishment, and geological framework. A total of 246 surface sand samples were collected from beach and foredune zones between 1993 and 2018. These samples were analyzed in relation to shoreline changes, hydrodynamic data, and geological context. The results show that sediment texture is most affected within 1–2 km downdrift and up to 4–5 km updrift of port structures. Downdrift areas tend to contain coarser, poorly sorted sediments because of erosion and the exposure of deeper strata, while updrift zones accumulate finer, well-sorted sands via longshore transport. In the long term, the geological framework controls sediment characteristics. In the medium term, introduced material that differs in grain size from natural beach sediments may alter the texture of the sediment, either coarsening or refining it. The latter slowly returns to its natural texture. Short-term changes are driven by storm events. These findings highlight the importance of integrating structural interventions, nourishment practices, and geological understanding for sustainable coastal management. Full article

Supercapacitors, with their remarkable attributes such as including a high power density, an extended cycle life, and inherent safety, have emerged as a major research area for electrochemical energy storage. In this paper, phenylenediamine and glyoxal were used as raw material to prepare p-phenylenediamine glyoxal (PGo) with one single step. p-phenylenediamine glyoxal-ferrocene (PGo-Fc_x, x = 1, 0.3, 0.2, 0.1) composites were synthesized based on a poly-Schiff base. FTIR and XRD results indicated that ferrocene doping preserves the intrinsic PGo framework while inducing grain refinement, as evidenced by the narrowing of the XRD diffraction peaks. SEM observations further revealed distinct morphological evolution. CV (cyclic voltammetry), EIS (electrochemical impedance spectroscopy), and GCD (galvanostatic charge–discharge) were conducted on PGo-Fc_x in order to examine its electrochemical performance. The PGo-Fc_0.3 in PGo-Fc_x electrode material had a specific capacitance of 59.6 F/g at a current density of 0.5 A/g and 36 F/g at a current density of 10 A/g. Notably, even after undergoing 5000 charging–discharging cycles at 10 A/g, the material retained 76.2% of its specific capacitance compared to the initial cycle. Therefore, taking conductive polymers and metal oxide materials for modification can improve the stability and electrochemical performance of supercapacitors. Full article

With the rapid development of sports analytics and artificial intelligence, accurate human pose estimation in badminton is becoming increasingly important. However, challenges such as the lack of domain-specific datasets and the complexity of athletes’ movements continue to hinder progress in this area. To address these issues, we propose an enhanced pose estimation framework tailored to badminton players, built upon an improved YOLOv8-Pose architecture. In particular, we introduce an efficient local attention (ELA) mechanism that effectively captures fine-grained spatial dependencies and contextual information, thereby significantly improving the keypoint localization accuracy and overall pose estimation performance. To support this study, we construct a dedicated badminton pose dataset comprising 4000 manually annotated samples, captured using a Microsoft Kinect v2 camera. The raw data undergo careful processing and refinement through a combination of depth-assisted annotation and visual inspection to ensure high-quality ground truth keypoints. Furthermore, we conduct an in-depth comparative analysis of multiple attention modules and their integration strategies within the network, offering generalizable insights to enhance pose estimation models in other sports domains. The experimental results show that the proposed ELA-enhanced YOLOv8-Pose model consistently achieves superior accuracy across multiple evaluation metrics, including the mean squared error (MSE), object keypoint similarity (OKS), and percentage of correct keypoints (PCK), highlighting its effectiveness and potential for broader applications in sports vision tasks. Full article

The decoding of motor imagery (MI) electroencephalogram (EEG) signals is crucial for motor control and rehabilitation. However, as feature extraction is the core component of the decoding process, traditional methods, often limited to single-feature domains or shallow time-frequency fusion, struggle to comprehensively capture the spatio-temporal-frequency characteristics of the signals, thereby limiting decoding accuracy. To address these limitations, this paper proposes a dual-branch neural network architecture with multi-domain feature fusion, the dual-branch spatio-temporal-frequency fusion convolutional network with Transformer (DB-STFFCNet). The DB-STFFCNet model consists of three modules: the spatiotemporal feature extraction module (STFE), the frequency feature extraction module (FFE), and the feature fusion and classification module. The STFE module employs a lightweight multi-dimensional attention network combined with a temporal Transformer encoder, capable of simultaneously modeling local fine-grained features and global spatiotemporal dependencies, effectively integrating spatiotemporal information and enhancing feature representation. The FFE module constructs a hierarchical feature refinement structure by leveraging the fast Fourier transform (FFT) and multi-scale frequency convolutions, while a frequency-domain Transformer encoder captures the global dependencies among frequency domain features, thus improving the model’s ability to represent key frequency information. Finally, the fusion module effectively consolidates the spatiotemporal and frequency features to achieve accurate classification. To evaluate the feasibility of the proposed method, experiments were conducted on the BCI Competition IV-2a and IV-2b public datasets, achieving accuracies of 83.13% and 89.54%, respectively, outperforming existing methods. This study provides a novel solution for joint time-frequency representation learning in EEG analysis. Full article

S355NL structural steel is extensively employed in bridges, ships, and power station equipment owing to its excellent tensile strength, weldability, and low-temperature toughness. However, pronounced fluctuations in its Charpy impact energy at low temperatures significantly compromise the reliability and service life of critical components. In this study, vacuum-induction-melted ingots of S355NL steel containing 0–0.086 wt.% rare earth cerium were prepared. The effects of Ce on microstructures, inclusions, and impact toughness were systematically investigated using optical microscopy (OM), scanning electron microscopy (SEM), electron backscatter diffraction (EBSD), and Charpy V-notch testing. The results indicate that appropriate Ce additions (0.0011–0.0049 wt.%) refine the average grain size from 5.27 μm to 4.88 μm, reduce the pearlite interlamellar spacing from 204 nm to 169 nm, and promote the transformation of large-size Al₂O₃-MnS composite inclusions into fine, spherical, Ce-rich oxysulfides. Charpy V-notch tests at –50 °C reveal that 0.0011 wt.% Ce enhances both longitudinal (269.7 J) and transverse (257.4 J) absorbed energies while minimizing anisotropy (E_t/E_l  =  1.01). Conversely, excessive Ce addition (0.086 wt.%) leads to coarse inclusions and deteriorates impact performance. These findings establish an optimal Ce window (0.0011–0.0049 wt.%) for microstructural and inclusion engineering to enhance the low-temperature impact toughness of S355NL steel. Full article

This study focuses on S32654 super austenitic stainless steel (SASS) and systematically characterizes the morphology of the sigma (σ) phase and the segregation behavior of alloying elements in its as-cast microstructure. High-temperature confocal scanning laser microscopy (HT-CSLM) was employed to investigate the effect of the rare earth element yttrium (Y) on the solidification microstructure and σ phase precipitation behavior of SASS. The results show that the microstructure of SASS consists of austenite dendrites and interdendritic eutectoid structures. The eutectoid structures mainly comprise the σ phase and the γ₂ phase, exhibiting lamellar or honeycomb-like morphologies. Regarding elemental distribution, molybdenum displays a “concave” distribution pattern within the dendrites, with lower concentrations at the center and higher concentrations at the sides; when Mo locally exceeds beyond a certain threshold, it easily induces the formation of eutectoid structures. Mo is the most significant segregating element, with a segregation ratio as high as 1.69. The formation mechanism of the σ phase is attributed to the solid-state phase transformation of austenite (γ → γ₂ + σ). In the late stages of solidification, the concentration of chromium and Mo in the residual liquid phase increases, and due to insufficient diffusion, there are significant compositional differences between the interdendritic regions and the matrix. The enriched Cr and Mo cause the interdendritic austenite to become supersaturated, leading to solid-state phase transformation during subsequent cooling, thereby promoting σ phase precipitation. The overall phase transformation process can be summarized as L → L + γ → γ → γ + γ₂ + σ. Y microalloying has a significant influence on the solidification process. The addition of Y increases the nucleation temperature of austenite, raises nucleation density, and refines the solidification microstructure. However, Y addition also leads to an increased amount of eutectoid structures. This is primarily because Y broadens the solidification temperature range of the alloy and prolongs grain growth perio, which aggravates the microsegregation of elements such as Cr and Mo. Moreover, Y raises the initial precipitation temperature of the σ phase and enhances atomic diffusion during solidification, further promoting σ phase precipitation during the subsequent eutectoid transformation. Full article

Wheat disease detection is a crucial component of intelligent agricultural systems in modern agriculture. However, at present, its detection accuracy still has certain limitations. The existing models hardly capture the irregular and fine-grained texture features of the lesions, and the results of spatial information reconstruction caused by standard upsampling operations are inaccuracy. In this work, the GDFC-YOLO method is proposed to address these limitations and enhance the accuracy of detection. This method is based on YOLOv11 and encompasses three key aspects of improvement: (1) a newly designed Ghost Dynamic Feature Core (GDFC) in the backbone, which improves the efficiency of disease feature extraction and enhances the model’s ability to capture informative representations; (2) a redesigned neck structure, Disease-Focused Neck (DF-Neck), which further strengthens feature expressiveness, to improve multi-scale fusion and refine feature processing pipelines; and (3) the integration of the Powerful Intersection over Union v2 (PIoUv2) loss function to optimize the regression accuracy and convergence speed. The results showed that GDFC-YOLO improved the average accuracy from 0.86 to 0.90 when the cross-overmerge threshold was 0.5 (mAP@0.5), its accuracy reached 0.899, its recall rate reached 0.821, and it still maintained a structure with only 9.27 M parameters. From these results, it can be known that GDFC-YOLO has a good detection performance and stronger practicability relatively. It is a solution that can accurately and efficiently detect crop diseases in real agricultural scenarios. Full article