Search Results (9,985)

Physics–Informed Neural Networks (PINNs) have demonstrated efficacy in solving both forward and inverse problems for nonlinear partial differential equations (PDEs). However, they frequently struggle to accurately capture multiscale physical features, particularly in regions exhibiting sharp local variations such as shock waves and discontinuities, and often suffer from optimization difficulties in complex loss landscapes. To address these issues, we propose LocRes–PINN, a physics–informed neural network framework that integrates local awareness mechanisms with residual learning. This framework integrates a radial basis function (RBF) encoder to enhance the perception of local variations and embeds it within a residual backbone to facilitate stable gradient propagation. Furthermore, we incorporate a residual–based adaptive refinement strategy and an adaptive weighted loss scheme to dynamically focus training on high–error regions and balance multi–objective constraints. Numerical experiments on the Extended Korteweg–de Vries, Navier–Stokes, and Burgers equations demonstrate that LocRes–PINN reduces relative prediction errors by approximately 12% to 67% compared to standard benchmarks. The results also verify the model’s robustness in parameter identification and noise resilience. Full article

With the continuous improvement of power system intelligence, multimodal data generated during distribution network maintenance have grown exponentially. However, existing power multimodal datasets commonly suffer from issues such as low sample quality, frequent factual errors, and inconsistent instruction expressions caused by regional differences.Traditional sample correction methods mainly rely on manual screening or single-feature matching, which suffer from low efficiency and limited adaptability. This paper proposes a multimodal sample correction framework based on large-model instruction enhancement and knowledge guidance, focusing on two critical modalities: temporal data and text documentation. Multimodal sample correction refers to the task of identifying and rectifying errors, inconsistencies, or quality issues in datasets containing multiple data types (temporal sequences and text), with the objective of producing corrected samples that maintain factual accuracy, temporal consistency, and domain-specific compliance. Our proposed framework employs a three-stage processing approach: first, temporal Bidirectional Encoder Representations from Transformers (BERT) models and text BERT models are used to extract and fuse device temporal features and text features, respectively; second, a knowledge-injected assessment mechanism integrated with power knowledge graphs and DeepSeek’s long-chain-of-thought (CoT) capabilities is designed to achieve precise assessment of sample credibility; third, beam search algorithms are employed to generate high-quality corrected text, significantly improving the quality and reliability of multimodal samples in power professional scenarios. Experimental results demonstrate that our method significantly outperforms baseline models across all evaluation metrics (BLEU: 0.361, ROUGE: 0.521, METEOR: 0.443, F1-Score: 0.796), achieving improvements ranging from 21.1% to 73.0% over state-of-the-art methods: specifically, a 21.1% improvement over GECToR in BLEU, 26.5% over GECToR in ROUGE, 30.3% over Deep Edit in METEOR, and 11.8% over Deep Edit in F1-Score, with a reduction of approximately 35% in hallucination rates compared to existing approaches. These improvements provide important technical support for intelligent operation and maintenance of power systems, with implications for improving data quality management, enhancing model reliability in safety-critical applications, and enabling scalable knowledge-guided correction frameworks transferable to other industrial domains requiring high data integrity. Full article

The incorporation of artificial intelligence (AI) and digital technology in healthcare has revolutionized service delivery, improving diagnostic precision, patient outcomes, and operational efficacy. Nonetheless, despite considerable progress, numerous problems persist that impede the realization of full potential. Current reviews predominantly emphasize the advantages of AI in disease detection and health guidance, neglecting significant concerns such as social opposition, regulatory frameworks, and geographical discrepancies. This SLR, executed in accordance with PRISMA principles, examined 21 publications from 2020 to 2025 to assess the present condition of AI and digital technologies inside Saudi Arabia’s healthcare industry. Initially, 863 publications were obtained, from which 21 were chosen for comprehensive examination. Significant discoveries encompass the extensive utilization of telemedicine, data analytics, mobile health applications, Internet of Things, electronic health records, blockchain technology, online platforms, cloud computing, and encryption methods. These technologies augment diagnostic precision, boost patient outcomes, optimize administrative procedures, and foster preventative medicine, contributing to cost-effectiveness, environmental sustainability, and enduring service provision. Nonetheless, issues include data privacy concerns, elevated implementation expenses, opposition to change, interoperability challenge, and regulatory issues persist as substantial barriers. Subsequent investigations must concentrate on the development of culturally relevant AI algorithms, the enhancement of Arabic natural language processing, and the establishment of AI-driven mental health systems. By confronting these challenges and utilizing emerging technologies, Saudi Arabia has the potential to establish its status as a leading nation in medical services innovation, guaranteeing patient-centered, efficient, and accessible healthcare delivery. Recommendations must include augmenting data privacy and security, minimizing implementation expenses, surmounting resistance to change, enhancing interoperability, fortifying regulatory frameworks, addressing regional inequities, and investing in nascent technologies. Full article

The classification of ship targets in remote sensing images holds significant application value in fields such as marine monitoring and national defence. Although existing research has yielded considerable achievements in ship classification, current methods struggle to distinguish highly similar ship categories for fine-grained classification tasks due to a lack of targeted design. Specifically, they exhibit the following shortcomings: limited ability to extract locally discriminative features; inadequate fusion of features at high and low levels of representation granularity; and sensitivity of model performance to background noise. To address this issue, this paper proposes a fine-grained classification framework for ship targets in remote sensing images based on Multi-Granularity Attention Network (MA-Net), specifically designed to tackle the aforementioned three major challenges encountered in fine-grained classification tasks for ship targets in remote sensing. This framework first performs multi-level feature extraction through a backbone network, subsequently introducing an Adaptive Local Feature Attention (ALFA) module. This module employs dynamic overlapping region segmentation techniques to assist the network in learning spatial structural combinations, thereby optimising the representation of local features. Secondly, a Dynamic Multi-Granularity Feature Fusion (DMGFF) module is designed to dynamically fuse feature maps of varying representational granularities and select key attribute features. Finally, a Feature-Based Data Augmentation (FBDA) method is developed to effectively highlight target detail features, thereby enhancing feature expression capabilities. On the public FGSC-23 and FGSCR-42 datasets, MA-Net attains top-performing accuracies of 93.12% and 98.40%, surpassing the previous best methods and establishing a new state of the art for fine-grained classification of ship targets in remote sensing images. Full article

The increasing frequency of extreme rainfall events that cause severe damage is considered a clear sign of climate change. Therefore, analyzing these events and gaining a better understanding of the circulation patterns that form precipitation regimes and trigger torrential rainfall are crucial for developing adaptation strategies. This study aims to present a comprehensive picture of precipitation regimes in Bulgaria under contemporary climate conditions, investigate the connections between precipitation and atmospheric circulation patterns, and propose a regionalization of torrential rainfall. We used daily precipitation data collected in the period 1991–2020, along with data on hazardous rainfall warnings issued by the National Institute of Meteorology and Hydrology. To identify the circulation patterns associated with both rainy days and hazardous rainfall in Bulgaria, we applied the automated Jenkinson–Collison classification. To identify precipitation patterns, we conducted a principal component analysis in T-mode with varimax rotation and k-means clustering of component scores on both monthly normals and a dataset of 166 selected torrential rainfall days. The results, examined in the context of the existing regionalization of precipitation, highlight the climatic diversity of precipitation regimes in Bulgaria. Our findings indicate that torrential rainfall is associated with low-pressure systems and airflows mainly from the east or northeast, as well as with weak-gradient pressure fields. Full article

Withcontinuous advancements in remote sensing technology, object detection in remote sensing images has emerged as a critical research direction in maritime surveillance, port management, and national defense. Among these applications, ship detection is a key task. Due to the fact that ships in images typically exhibit arbitrary rotations, multi-scale distributions, and complex backgrounds, conventional detection methods based on horizontal or rotated bounding boxes often fail to adequately capture the fine-grained information of the targets, thereby compromising detection accuracy. This paper proposes the Dense Representative Points-Guided Rotated-Ship Detection (DenseRRSD) method. The proposed approach represents ship objects using dense representative points (RepPoints) to effectively capture local semantic information, thereby avoiding the background noise issues associated with traditional rectangular bounding box representations. To further enhance detection accuracy, an edge region sampling strategy is devised to uniformly sample RepPoints from critical ship parts, and a Weighted Residual Feature Pyramid Network (WRFPN) is introduced to efficiently fuse the multi-scale features through residual connections and learnable weights. In addition, a Weighted Chamfer Loss (WCL) combined with a staged localization loss strategy is employed to progressively refine localization from coarse to fine stages. Experimental results on both the HRSC2016 dataset and the newly constructed DOTA-SHIP dataset demonstrate that DenseRRSD achieves state-of-the-art detection accuracy, with mean Average Precision (mAP) scores of 91.2% and 83.2%, respectively, significantly outperforming existing methods. These results verify the effectiveness and robustness of the proposed approach in rotated-ship detection under diverse conditions. Full article

Multi-axle vehicles are commonly used for heavy-duty special operations, which easily leads to high driving torque demands when adopting distributed electric drive configurations. This study achieves the objective of reducing the driving torque of each in-wheel motor while controlling the stability of multi-axle vehicles. Taking a five-axle distributed drive test vehicle as the research object, a hierarchical control strategy integrating active all-wheel steering and direct yaw moment control is proposed. The upper layer is implemented based on model predictive control, with fuzzy control introduced to dynamically adjust control weights; the lower layer accomplishes the allocation of targets calculated by the upper layer through minimizing the objective function of tire load ratio. A linear parameter varying (LPV) tire model is introduced into the vehicle model to improve the calculation accuracy of tire lateral forces, and a neural network method is employed to solve the real-time performance issue of the model predictive control (MPC) controller. The proposed strategy is verified through a combination of simulation and real vehicle tests. High-speed condition simulations demonstrate that the AWS/DYC strategy significantly outperforms the ARS/DYC approach: compared to the active rear-wheel steering strategy, while the sideslip angle is reduced by 90.98%, the peak driving torque is reduced by 30.78%. Notably, tire slip angle analysis reveals that AWS/DYC maintains relatively uniform slip angle distribution across axles with a maximum of 4.7°, entirely within the linear working region, optimally balancing tire performance utilization with lateral stability while preserving safety margin, whereas ARS/DYC causes slip angles to exceed 11.9° at the rear axle, entering saturation. Low-speed real vehicle tests further confirm the engineering applicability of the strategy. The proposed method is of significant importance for the application of distributed drive configurations in the field of special vehicles. Full article

Timely and accurate estimation of regional winter wheat yield is of great significance for safeguarding food security and promoting sustainable agricultural development. In recent years, deep learning has been widely applied in crop yield estimation due to its powerful capability in mining complex relationships. However, the irregular shapes of administrative regions pose challenges for integrating spatial data such as remote sensing into deep learning models. To address this issue, this study employed mean-based aggregation and histogram-based dimensionality reduction techniques to preprocess spatial data, including remote sensing and meteorological data, thereby generating sample sets suitable for deep learning models. This study identified the phenomenon of feature conflict when processing heterogeneous features in conventional Long Short-Term Memory (LSTM) models and proposed a TB-LSTM (Two-Branch LSTM) model to mitigate such conflicts. The impact of different input feature combinations on estimation accuracy was analyzed, and the model’s capability for early yield prediction was further evaluated. The results show that: (1) The proposed TB-LSTM model achieved superior performance (R²: 0.853, RMSE: 516.619 kg/ha) compared to the baseline LSTM (R²: 0.353–0.732; RMSE: 735.378–1126.062 kg/ha), confirming its efficiency in resolving feature conflict and better exploiting the yield estimation potential of remote sensing and meteorological data. (2) The integration of meteorological data, spectral reflectance, and vegetation indices proved essential for achieving optimal yield estimation accuracy. Meteorological data provided the most significant contribution, while spectral reflectance and vegetation indices offered complementary information that improved model robustness. When all three data types were utilized simultaneously, the TB-LSTM model achieved peak estimation accuracy (R²: 0.853; RMSE: 514.013 kg/ha; MAE: 380.563 kg/ha). (3) The TB-LSTM model demonstrated robust early prediction capability. Using data from the first 27 time phases (covering growth stages up to heading), it successfully predicted winter wheat yields 48 days before harvest with optimal precision (R²: 0.868; RMSE: 487.327 kg/ha; MAE: 361.353 kg/ha). This capability supports proactive decision-making and resource allocation in agricultural management. Full article

Power-line presence recognition technology for unmanned aerial vehicles (UAVs) is one of the key research directions in the field of UAV remote sensing. With the rapid development of UAV technology, the application of UAVs in various fields has become increasingly widespread. However, when flying in urban and rural areas, UAVs often face the danger of obstacles such as power lines, posing challenges to flight safety and stability. To address this issue, this study proposes a novel method for presence recognition in UAVs for power lines using an improved GhostNet v2 knowledge distillation dual-attention mechanism convolutional neural network. The construction of a real-time UAV power-line presence recognition system involves three aspects: dataset acquisition, a novel network model, and real-time presence recognition. First, by cleaning, enhancing, and segmenting the power-line data collected by UAVs, a UAV power-line presence recognition image dataset is obtained. Second, through comparative experiments with multi-attention modules, the dual-attention mechanism is selected to construct the CNN, and the UAV real-time power-line presence recognition training is conducted using the SGD optimiser and Hard-Swish activation function. Finally, knowledge distillation is employed to transfer the knowledge from the dual-attention mechanism-based CNN to the nonlinear function and Mamba-enhanced GhostNet v2 network, thereby reducing the model’s parameter count and achieving real-time recognition performance suitable for mobile device deployment. Experiments demonstrate that the UAV-based real-time power-line presence recognition method proposed in this paper achieves real-time recognition accuracy rates of over 91.4% across all regions, providing a technical foundation for advancing the development and progress of UAV-based real-time power-line presence recognition. Full article

Electric vehicles (EVs) and renewable energy generation are widely regarded as key drivers of low-carbon transformation in the transportation and energy sectors due to their emission reduction potential and environmental benefits. However, the inherent intermittency and volatility of photovoltaic (PV) power, coupled with increasingly stochastic and disorderly EV charging demand, pose significant challenges to grid stability and local renewable energy utilization. To address these issues, this paper proposes a dynamic pricing optimization approach based on a Stackelberg game framework, in which the PV charging station operator acts as the leader and EV users as followers. Unlike conventional models, the proposed framework explicitly incorporates user psychological expectations and response deviations through a three-stage “dead-zone-linear-saturation” responsiveness structure, thereby capturing the uncertainty and partial rationality of EV charging behavior. The upper-level objective seeks to maximize operator profit and enhance PV self-consumption, while the lower-level objective minimizes user energy cost under price-responsive charging decisions. The bilevel optimization problem is solved via a differential evolution (DE) algorithm combined with YALMIP+CPLEX. Simulation results for a regional PV-EV charging station show that the proposed strategy increases PV self-consumption to about 90.5% and shifts the load peak from 18:00–20:00 to 10:00–15:00, effectively aligning charging demand with PV output. Compared with both flat and standard time-of-use (TOU) tariffs, the dynamic pricing scheme yields higher operator profit (about 7% improvement over flat pricing) while keeping total user energy expenditure essentially unchanged. In addition, the cumulative carbon reduction cost over the operating cycle is reduced by approximately 4.1% relative to flat pricing and 1.9% relative to TOU pricing, demonstrating simultaneous economic and environmental benefits of the proposed game-based dynamic pricing framework. Full article

Snakebite envenoming remains a critical public health issue, and the molecular variability of venoms limits the cross-species efficacy of conventional antivenoms. Here, we conducted a comparative proteomic analysis of Bitis arietans venom to identify conserved peptide regions derived from enzymatic toxins and evaluate their potential relevance for complementary immunotherapeutic applications. Enzyme-enriched venom fractions were isolated through sequential affinity and ion-exchange chromatography and were subsequently characterized using fluorogenic FRET substrates and inhibitor assays. LC–MS/MS analysis identified 1099 proteins and revealed 36 conserved peptides within snake venom metalloproteinases (SVMPs), serine proteases (SVSPs), and phospholipase A₂ (PLA₂), particularly located near catalytic residues and structurally essential motifs such as the HExxHxxGxxH zinc-binding site in SVMPs, the His-Asp-Ser catalytic triad in SVSPs, and the Ca²⁺-binding loop in PLA₂, across Viperidae venoms. These conserved regions were also observed in homologous toxin isoforms from additional Viperidae genera, supporting the evolutionary conservation of key functional domains. While sequence conservation alone does not guarantee neutralization capacity, the identified regions represent strong candidates for structural epitope mapping and targeted antibody development. This study provides a peptide-level framework for advancing complementary antibody-based therapies designed to broaden cross-species toxin recognition, reduce antivenom dosage requirements, and improve clinical outcomes in snakebite envenoming. Full article

Constructing ecological networks (ENs) is an effective measure to mitigate the conflict between urban development and ecological conservation. However, existing simulating methods lack adequate consideration of human ecological demands and the spatial scale differences between these demands and natural ecological processes. This might lead to issues such as incomplete ecological process cycles or structural mismatches being overlooked during ENs simulations. To address these gaps, this study proposed an urban multi-scale nested ENs simulating framework that integrates human ecological demands with natural ecological processes. The framework first simulated an ENs focused on natural ecological process cycles at a global scale (GS). Then, it simulated an ENs centered on human ecological needs within the core urban areas at local scale (LS). Finally, it nested these multi-scale ENs by using cross-scale ecological supply sources as connecting points. This framework was applied to simulate spatio-temporal pattern changes in ENs of Jinan City, a core city in downstream of the Yellow River in China, aiming to mitigate cross-scale ecological conflicts between human–nature interactions under the background of urbanization. The study’s findings revealed that the area of demand sources increased by 8.56 times over 20 years. the area of cross-scale supply sources decreased by 15 km² relative to 2000, and the deterioration in connectivity was more pronounced in GS compared to LS, with a decline of approximately 13.8%. These changes indicate the presence of incomplete ecological process cycles and structural mismatches across the multi-scale boundaries within the study area, which have been worsening annually. We recommend optimizing Jinan City’s multi-scale ecological network through three key strategies: rectifying internal structural mismatches, protecting core ecological areas, and aligning regional ecological demands. Implementing these strategies could significantly improve the network structure, reduce cross-scale mismatches, and enhance ecological connectivity by about 9%. This study highlights the importance of addressing structural mismatches and promoting complete ecological cycles in urban multi-scale ENs simulating, providing valuable insights for formulating urban multi-scale ecological conservation and restoration policies. Full article

Underwater imaging suffers from significant degradation due to scattering by suspended particles, selective absorption by the medium, and depth-dependent noise, leading to issues such as contrast reduction, color distortion, and blurring. Existing enhancement methods typically address only one aspect of these problems, relying on unrealistic assumptions of uniform noise, and fail to jointly handle the spatially heterogeneous noise and spectral channel attenuation. To address these challenges, we propose the variational-based spatial–spectral joint enhancement method (VSJE). This method is based on the physical principles of underwater optical imaging and constructs a depth-aware noise heterogeneity model to accurately capture the differences in noise intensity between near and far regions. Additionally, we propose a channel-sensitive adaptive regularization mechanism based on multidimensional statistics to accommodate the spectral attenuation characteristics of the red, green, and blue channels. A unified variational energy function is then formulated to integrate noise suppression, data fidelity, and color consistency constraints within a collaborative optimization framework, where the depth-aware noise model and channel-sensitive regularization serve as the core adaptive components of the variational formulation. This design enables the joint restoration of multidimensional degradation in underwater images by leveraging the variational framework’s capability to balance multiple enhancement objectives in a mathematically rigorous manner. Experimental results using the UIEBD-VAL dataset demonstrate that VSJE achieves a URanker score of 2.4651 and a UICM score of 9.0740, representing a 30.9% improvement over the state-of-the-art method GDCP in the URanker metric—a key indicator for evaluating the overall visual quality of underwater images. VSJE exhibits superior performance in metrics related to color uniformity (UICM), perceptual quality (CNNIQA, PAQ2PIQ), and overall visual ranking (URanker). Full article

The quality of tea bud harvesting directly affects the final quality of the tea; however, due to the small size of tea buds and the complex natural background, accurately detecting them remains challenging. To address this issue, this paper proposes a lightweight and efficient tea bud detection model named TSF-Net. This model adopts the P2-enhanced bidirectional feature pyramid network (P2A-BiFPN) to enhance the recognition ability of small objects and achieve efficient multi-scale feature fusion. Additionally, coordinate space attention (CSA) is embedded in multiple C3k2 blocks to enhance the feature extraction of key regions, while an A2C2f module based on self-attention is introduced to further improve the fine feature representation. Extensive experiments conducted on the self-built WYTeaBud dataset show that TSF-Net increases mAP@50 by 2.0% and reduces the model parameters to approximately 85% of the baseline, achieving a good balance between detection accuracy and model complexity. Further evaluations on public tea bud datasets and the VisDrone2019 small object benchmark also confirm the effectiveness and generalization ability of the proposed method. Moreover, TSF-Net is converted to the RKNN format and successfully deployed on the RK3588 embedded platform, verifying its practical applicability and deployment potential in intelligent tea bud harvesting. Full article

For a long time, low efficiency in the transfer of rural collective land use rights and the ambiguous attribution of collective land property rights have not only restricted the mobility of rural labor factors but have also hindered the release of vitality in the rural collective economy. This has resulted in lagging growth in the income that rural residents obtain from collective economic factors, contributing to the persistent widening of the urban/rural income gap. As an important institutional innovation to address these issues, the effects of the reform of the rural collective property rights system urgently need to be clarified. The reform of the rural collective property rights system constitutes a major initiative in the transformation of the rural land system. Centered on asset verification and valuation, as well as the demarcation of membership rights and the restructuring towards a shareholding cooperative system, it aims to establish a collective property rights regime characterized by clearly defined ownership and fully functional entitlements. This study takes the national pilot reform of rural collective property rights launched in 2016 as a quasi-natural policy experiment, systematically examining the impact of this pilot policy on the internal income gap within households and its spillover effects on the urban–rural income gap. Based on microdata from the China Household Finance Survey (CHFS) and the China Longitudinal Night Light Data Set (PANDA-China), this study constructs a five-period balanced panel dataset covering 2304 rural households across 25 provinces. A relative exploitation index based on the Kawani index is constructed, and empirical analysis is conducted using a combination of multi-period difference-in-differences (Multi-period DID), discrete binary models, and propensity score matching-difference-in-differences (PSM-DID) models. The results show that: First, the pilot reform significantly reduced the level of income inequality within rural areas in the pilot regions, and its policy benefits further generated positive spillovers via market-driven factor allocation mechanisms, effectively bridging the urban–rural income gap. Second, institutional reforms activated the potential of rural non-agricultural economic factors, establishing new channels for a two-way flow of urban and rural factors, becoming an important path to achieve the goal of common prosperity. Third, the policy effects exhibited significant heterogeneity, specifically manifested in the attributes of major grain-producing regions, initial household income levels, and the human capital characteristics of household heads having significant moderating effects on reform outcomes. This study not only provides theoretical support and empirical evidence for deepening rural property rights reforms under the new rural revitalization strategy, but it also reveals the driving role of institutional innovation in factor mobility, thereby influencing the transmission mechanism of income distribution patterns. This finding offers a China-based solution for developing countries to address the imbalance in urban–rural development and the widening income gap. Full article