Search Results (12,637)

Wind power forecasting plays a crucial role in power dispatch and safety management of wind farms. However, the insufficient integration of multimodal heterogeneous data and the limitations of conventional graph construction strategies significantly restrict forecasting performance. Existing approaches either rely on simple feature aggregation, which cannot fully capture cross-modal dependencies, or adopt predefined or single-criterion graph construction methods that fail to characterize complex turbine relationships involving spatial, temporal, and nonlinear correlations. To address these challenges, this paper proposes a Multimodal Adaptive Fusion Graph Neural Network (MAF-GNN) for short-term wind power forecasting. First, a Modality-Aware Representation Learning (MARL) module is developed to extract informative multimodal representations by modeling modality-specific characteristics and cross-modal dependencies through attention-based fusion. Second, an Adaptive Graph Learning with Multi-Similarity (AGL-MS) module is introduced to parametrically integrate four complementary similarity priors—geographic distance, Dynamic Time Warping (DTW), Maximal Information Coefficient (MIC), and cosine similarity—for adaptive turbine correlation graph construction. Furthermore, a Credibility-Modulated Graph Convolutional Network (CM-GCN) is developed to reduce the influence of unreliable node information during message propagation. Extensive experiments conducted on the SDWPF dataset demonstrate that MAF-GNN reduces MAE by 14.0–21.3% compared with sequential baselines and achieves 5.3–10.2% improvement over spatiotemporal graph-based models. Ablation studies further verify the complementary effectiveness of each proposed module in improving forecasting performance. Full article

Surface water quality is critical to ecosystem health and sustainable development, yet conventional monitoring falls short of spatiotemporally continuous assessment. Remote sensing coupled with machine learning has become a powerful paradigm for large-scale quantitative retrieval of water quality parameters (WQPs). This review examines the progress and prospects of machine-learning-driven multi-source remote sensing for surface WQP retrieval. A systematic literature review following PRISMA 2020 guidelines, covering 437 Web of Science Core Collection publications (2000–2025), reveals exponential growth, with China and the United States contributing 70.3% of total output. A critical synthesis covers four dimensions: (1) characteristics and fusion strategies of satellite, airborne, and ground-based remote sensing data; (2) modeling features of traditional machine learning (SVR, RF, GBDT), deep learning (CNN, RNN, Transformer), and hybrid approaches; and (3) retrieval advances for optically active versus non-optically active parameters—the former approaches operational readiness while the latter remains constrained by weak indirect spectral correlations; and (4) uncertainty sources and mitigation strategies across the data–model–parameter chain. Five key challenges are identified: limited model generalizability, insufficient physical interpretability, optical heterogeneity and parameter coupling, scarce in situ data, and multi-source fusion bottlenecks. Five future directions are proposed—transfer learning, physically informed explainable machine learning, non-optically active parameter retrieval, benchmark dataset development, and intelligent multi-source fusion—offering a roadmap toward operational surface water quality monitoring. Full article

The output of photovoltaic power generation exhibits significant randomness, volatility, and intermittency, and its uncertainty will have an impact on the planning assessment, dispatch decision-making, and real-time operation of the power system. To systematically understand the development status of modeling methods for the uncertainty of photovoltaic power generation, this paper conducts a review around the generation of annual scenarios and multi-timescale power prediction of photovoltaic power, and analyzes the correlations between photovoltaic output, influencing factors, and system applications from the perspective of statistical relationships and artificial intelligence. For the annual scale, the focus is on the generation methods of meteorological-driven scenarios for long-term sequences, including probability statistical methods, deep generation methods, constraint relations and engineering application evaluation issues; for the day-ahead scale, the historical power, meteorological variables and numerical weather forecasts are used to explore feature extraction, probability prediction and robust modeling methods; for the intraday scale, the signal decomposition, deep learning, regional collaborative modeling and multi-source perception methods for short-term power fluctuations perception are summarized. On this basis, further analysis is conducted on subsequent research directions such as multi-timescale collaborative modeling, multi-source heterogeneous information fusion, controllable generative modeling, extreme scenario characterization, and engineering closed-loop verification. This paper can serve as a reference for scenario generation, power prediction, and power system operation analysis under the condition of a high proportion of photovoltaic power integration. Full article

Semantic segmentation of 3D point clouds remains difficult when LiDAR or depth-camera data are sampled unevenly. This paper presents DANet, a 3D semantic segmentation framework built on joint density- and semantics-adaptive convolution. Its core operator, Density-Adaptive Radius Convolution (DAR-Conv), predicts point-wise neighborhood radii before feature aggregation by combining density-driven initialization with semantics-aware modulation. In this way, dense regions can use compact receptive fields, whereas sparse or semantically complex regions can draw on broader contextual support. DANet also includes a Gated Adaptive Cross-Layer Fusion (GACF) module, which aligns encoder–decoder features and performs gated fusion with residual refinement. Experiments on S3DIS and NPM3D show that DANet obtains the highest reported mean accuracy (mAcc) among the compared methods on S3DIS, and high mean Intersection over Union (mIoU) and overall accuracy (OA) on NPM3D, supporting the usefulness of density- and semantics-aware receptive-field adaptation. Full article

Tomato leaf disease detection in natural environments is challenged by subtle early-stage symptoms, complex backgrounds, leaf occlusion, and scale variation, which can lead to missed detections, false detections, and unstable leaf localization. Meanwhile, practical agricultural applications impose higher requirements on model lightweightness and edge-deployment capability. To address these issues, this study proposes DFR-YOLOv12n, a lightweight tomato leaf disease detection model based on YOLOv12n that integrates detail-preserving downsampling, feature enhancement, and regression optimization. First, a multi-source dataset collected in natural environments was constructed and curated, covering eight categories: bacterial spot, early blight, late blight, leaf mold, mosaic virus disease, septoria leaf spot, yellow leaf curl virus disease, and healthy leaves. Second, SPDConv was introduced into key downsampling layers to preserve fine-grained disease-related visual cues. The A2C2f_DEConv module was incorporated into the P3 feature fusion branch to enhance leaf texture and disease-related appearance features under complex backgrounds. In addition, MPDIoU was adopted to optimize bounding box regression and improve whole-leaf localization under occlusion and background interference. The optimal model configuration was determined through insertion-position, module comparison, and ablation experiments. Compared with the baseline model, DFR-YOLOv12n increased Precision, Recall, and mAP@0.5 from 86.8%, 76.9%, and 86.5% to 88.1%, 81.7%, and 88.6%, respectively. Meanwhile, FLOPs decreased from 5.83 G to 5.27 G, the parameter count decreased from 2.51 M to 2.25 M, and the model size decreased from 5.22 MB to 4.71 MB. Furthermore, the model was successfully deployed and validated on the Jetson Nano platform, demonstrating its potential for edge applications. The results indicate that DFR-YOLOv12n achieves a favorable balance among detection accuracy, model complexity, and deployment feasibility, providing a reference for intelligent tomato leaf disease detection in natural environments. Full article

High-precision and resilient indoor positioning is a fundamental requirement for the autonomous operation of mobile robots in GNSS-denied environments. While visual sensors are commonly used for odometry, their operational reliability can be compromised in challenging scenarios such as drastic illumination fluctuations and sparse-textured environments. To address these sensor limitations, this study presents MVO, a magneto-visual odometry framework that explores indoor magnetic field anomalies as complementary constraints for visual odometry. By integrating a 30-magnetometer planar array model with a stereo camera, the proposed system establishes a multi-modal perception framework for indoor spaces. In the frontend, magnetic field gradient information is utilized to provide relative-pose constraints, which assist in the matching of image feature points and help maintain tracking continuity under visual degradation. In the backend, a factor graph optimization (FGO) framework incorporates magnetic relative-pose factors and visual reprojection factors into a unified optimization objective, which is then solved using the incremental smoothing and mapping 2 (iSAM2) algorithm. Frontend-level simulations are conducted to analyze the effects of magnetometer spatial configuration, sensor number, and calibration-error sensitivity on magnetic relative-pose estimation and covariance. Trajectory-level evaluations are further performed on the EuRoC dataset augmented with high-fidelity synthesized magnetic field data, including localization accuracy and computational load. Under this synthesized magnetic field setting, MVO shows improved localization accuracy and moderate computational load compared with the selected MSCKF-Stereo and VINS-Fusion reference baselines. These results provide a simulation-based feasibility validation of integrating magnetic field constraints with visual information for indoor odometry, while validation with real magnetometer array measurements remains future work. Full article

Sleep staging based on PSG is largely confined to clinical settings, while home-based sleep monitoring often faces the challenges of insufficient unimodal information and missing modalities. Aiming to overcome these challenges, this paper proposes a unified multimodal model for sleep staging based on cardiopulmonary signals. First, a heterogeneous multi-scale feature encoder with long and short branches is adopted to adapt to the cross-modal heterogeneity of ECG and THX. It combines a Transformer encoder and a Dilated CNN to complete feature fusion and temporal modeling. Subsequently, the unified model adaptively handles flexible modality combinations by introducing global context via a modal feature alignment strategy, which is built upon a framework consisting of a bimodal global branch and unimodal dedicated branches. On the SHHS dataset, the proposed model achieved Cohen’s kappa coefficients of 0.7547, 0.7121, and 0.7305 for four-stage sleep classification under ECG+THX, ECG-only, and THX-only inputs, respectively, demonstrating consistent improvements over three separately trained individual models. Furthermore, the model exhibits robust generalization performance on the P2018 external dataset and across samples with different severity levels of SDB. This work establishes a reliable algorithmic baseline for unobtrusive, long-term home sleep monitoring with missing modalities. Full article

Rotating machinery has been widely used in industries, but it often faces a high incidence of sudden failures under harsh operating conditions. Therefore, ensuring its safety and reliability is of utmost importance. However, fault diagnosis frequently encounters challenges such as limited training samples and the susceptibility of individual vibration sensors to external interference and noise. To enhance the recognition accuracy of rotating machinery under noisy and small-sample conditions, a fault diagnosis method for small samples based on multi-sensor fusion and CWT-CNN-BiLSTM is proposed in the paper. Firstly, the data from the multi-sensor is concatenated and fused, and then converted into a two-dimensional feature image through CWT. This study introduces a CNN-BiLSTM model designed to extract pivotal features from images. The feasibility of this method has been verified using the rotating machinery fault diagnosis database made available by the Korean Institute of Science and Technology. The average diagnostic accuracy achieved is 99.90%. The experimental results show that the proposed method results in more accurate and robust fault classification under small-sample conditions. Full article

Egocentric multi-view image analysis refers to the processing of utilizing synchronized video streams captured from multiple wearable cameras worn on the head or body, providing complementary first-person perspectives of dynamic, real-world interactions. Unlike single-view egocentric vision, which may suffer from severe occlusions, motion blur, and limited field-of-view or traditional fixed-camera multi-view setups (assuming static geometry and controlled environments), egocentric multi-view systems leverage body-worn rigs to enable a more robust and flexible 3D understanding in open-world, mobile scenarios. In this work, we present a systematic survey of advancements in cross-view feature fusion, geometric consistency enforcement, open-world detection, human–object interaction (HOI) modeling, action segmentation, 3D reconstruction, and novel-view synthesis specifically tailored to wearable multi-camera platforms. Key datasets released between 2024 and 2026—including HOT3D (833 min of synchronized multi-view hand/object interactions from Project Aria and Quest 3), MultiEgo (first multi-egocentric dataset for 4D social scene reconstruction), and Ego-1K (large-scale 12-camera rig for dynamic 3D video synthesis) are thoroughly examined alongside an analysis of integrations with large language models (LLMs) and vision–language models that drive performance gains, typically in the 15–30% range over single-view baselines in hand tracking, HOI recognition, and reconstruction fidelity, although we show through a consolidated meta-analysis that this gain is task-dependent: larger for geometry-bottlenecked tasks such as in-hand object lifting, and smaller, method-dependent, or occasionally negative for semantic-recognition tasks such as keystep recognition under naive view fusion. These methods cover work in multi-view stereo, cross-view learning, and novel-view synthesis while addressing several real-time wearable constraints. Practical applications such as immersive Augmented Reality/Virtual Reality (AR/VR), assistive robotics, and healthcare monitoring are also discussed together with the challenges in motion calibration, benchmark diversity, and edge deployment ability. Thus, in this review, we attempt to fill a critical gap by focusing exclusively on wearable multi-view systems in an open-world setting, synthesizing the latest literature to chart future directions toward more embodied and continual learning agents. Full article

High-speed maneuvering flight vehicles operating in the subsonic-to-transonic regime (250–500 m/s) pose severe challenges to defense interception systems due to their rapid and unpredictable maneuvering behaviors. Accurate short-term trajectory prediction is essential for effective terminal-phase interception guidance. This paper proposes DSGF-Net (Dual-Stream Gated Fusion Network), a hybrid deep learning architecture for 3D trajectory prediction that simultaneously exploits frequency-domain and temporal-domain information through independent parallel streams. DSGF-Net employs two Temporal Convolutional Networks (TCNs) as parallel encoders: a frequency stream processes Wavelet Packet Decomposition (WPD) features (24-dimensional, db4 wavelet, level-3 decomposition), and a temporal stream processes raw 3D coordinates. An adaptive sigmoid gating module dynamically fuses the two independently encoded streams at each time step and feature dimension, followed by an LSTM sequence learner and a single-step fully connected decoder. Experiments on a simulated dataset covering five representative maneuvering modes (cruise, dive, climb, serpentine, composite) reveal a three-level performance hierarchy. First, incorporating raw 3D coordinates alongside WPD features substantially improves clean-data accuracy over WPD-only baselines: DSGF-Net achieves ADE = 3.476 ± 0.010 m (5 random seeds) versus TCN-LSTM (WPD-only) at 3.938 ± 0.103 m (11.7% improvement). Second, a single-stream concatenation baseline (Concat-TCNLSTM) using identical inputs achieves comparable clean-data accuracy (3.333 ± 0.009 m), confirming that input information—rather than fusion mechanism—drives clean-data gains. Third, and most critically, DSGF-Net’s independently encoded dual-stream architecture enables adaptive suppression of degraded sensor inputs: under multi-sensor noise (complementary radar/GPS profiles), DSGF-Net achieves ADE = 13.91 m versus TCN-LSTM’s 21.32 m (34.9% advantage)—a substantially larger margin than on clean data—a capability structurally unavailable to concatenation-based models. With 300K parameters and a 4.96 ms inference time on an A100 GPU, DSGF-Net meets real-time terminal interception requirements (<10 ms). Full article

Marine monitoring records collected from buoys and nearshore sensors are often affected by missing values, abrupt spikes, and short-term fluctuations. These errors are difficult to remove with a single detection or interpolation rule, especially when local anomalies and global outliers occur in the same sequence. This study develops a Multi-Strategy Fusion Architecture (MSFA) for cleaning marine environmental time-series data. In MSFA, DBSCAN is not applied directly to the raw observations; instead, the time index and measurement value are first normalized into a common feature space, where local density anomalies can be detected more consistently. IQR screening is then used to identify global extreme values. After abnormal positions are marked, the repair result is estimated from two complementary sources: linear interpolation, which follows local temporal change, and a moving average based only on neighboring valid observations, which reduces random noise. Their contributions are adjusted according to local reliability rather than fixed manually. Because initial repair may still leave small residual errors, we further use a Combined Residual Metric (CRM) with a median/MAD-based threshold to recheck the repaired sequence and update the abnormal-position set when necessary. Experiments on the 2020 Dongying offshore buoy dataset and a self-collected nearshore dataset show that MSFA achieves AUROC/AUPRC/NRMSE values of 0.896/0.855/0.066 and 0.986/0.915/0.0653, respectively. Compared with DBSCAN+LOF, DBSCAN+Transformer, and IQR+Sigmoid, MSFA improves AUROC and AUPRC by about 12–25% on average and reduces NRMSE by more than 40%. These results indicate that the proposed method can improve the usability of noisy and incomplete marine monitoring data while keeping the cleaning process interpretable. Full article

Quality deterioration of chilled lamb during storage poses a challenge to meat quality and safety control, making rapid and accurate freshness-grade classification essential. Existing methods based on either spectral information or RGB image information alone are insufficient to simultaneously characterize internal chemical changes and external appearance changes during lamb quality deterioration. To address this issue, this study developed a chilled lamb freshness-grade classification method by integrating deep features from two-dimensional visible–near-infrared (Vis–NIR) spectral encoding with RGB image features. In this method, one-dimensional Vis–NIR spectra were transformed into two-dimensional encoded images using Gramian angular difference field (GADF), Gramian angular summation field (GASF), Markov transition field (MTF), and recurrence plot (RP) to enhance the representation of inter-wavelength structural relationships in spectral sequences, thereby compensating for the limited ability of conventional one-dimensional spectral modeling to capture global correlations and local variation information. Meanwhile, recursive feature elimination (RFE)-selected spectral deep features were fused with Spearman-selected RGB image features to construct a deep–shallow classification model. The results showed that the fusion models outperformed the single-modality models, with GADF(10%)+Image-SVM achieving the best performance, yielding an accuracy, F1-score, and MCC of 0.966, 0.957, and 0.946, respectively. Shapley additive explanations (SHAP) analysis further indicated that GADF deep features were the primary contributors, while RGB image features provided effective complementary information, demonstrating the potential of the proposed method for rapid and nondestructive freshness-grade classification of chilled lamb. Full article

Underwater fish detection (UFD) is pivotal for smart aquaculture and marine ecological monitoring. While recent detectors improve accuracy using stacked feature extractors or heavy attention modules, they often incur substantial computational overhead and, more importantly, neglect the underwater optical degradation that fundamentally limits UFD: wavelength-dependent absorption and turbidity-induced scattering significantly degrade contrast, blur fine structures, and introduce back-scattering noise, leading to unreliable localization and recognition. To address these challenges, we propose FinSight-Net, an efficient and prior-guided detection framework tailored to complex aquaculture environments. FinSight-Net introduces a Prior-guided Optical Decoupling (PIOD) bottleneck for degradation-aware underwater feature representation. PIOD decomposes features into four degradation-aware branches—forward-scattering blur, back-scattering veil, spectral attenuation, and high-frequency structural decay—and supervises their degradation response maps using image-derived pseudo priors. We further design a Sparsity-constrained High-frequency Adaptive Routing Pyramid (SHARP-FPN), which uses gradient-based salience responses and sparse cross-scale fusion to preserve boundary-sensitive cues and reduce redundant skip contributions, improving fish localization under severe blur and turbidity. Extensive experiments on DeepFish, AquaFishSet, and our challenging UW-BlurredFish benchmark demonstrate that FinSight-Net achieves state-of-the-art performance. In particular, on UW-BlurredFish, FinSight-Net achieves a state-of-the-art 92.8% mAP${}_{50}$, outperforming YOLOv11s, YOLOv12s, and YOLO26s by 4.8%, 2.0%, and 1.8%, respectively, while reducing the parameter count by 28.7%, 28.0%, and 29.5%, providing a strong and lightweight solution for real-time automated monitoring in smart aquaculture. Full article

Autonomous Underwater Vehicles (AUVs) face significant challenges in path-following control due to strong environmental disturbances and model uncertainties. To address these issues, this paper proposes a model-free deep reinforcement learning framework, named ILLT (Improved LOS-LSTM-TD3), which integrates an integral line-of-sight (LOS) guidance law with the twin delayed deep deterministic policy gradient (TD3) algorithm. The framework treats the LOS look-ahead distance as a learnable optimization variable and incorporates an LSTM network to capture temporal motion dependencies. A progressive unfreezing transfer learning strategy, combined with attention-based feature–current fusion, is designed to enhance domain adaptation under varying ocean currents. Simulation results demonstrate that ILLT reduces the average cross-track error by 48.5% compared to the baseline ILT algorithm and by 66.4% compared to traditional PID control, while achieving significantly faster convergence in target domains. Physical experiments in tank and lake environments further validate the algorithm’s feasibility and robustness, with tracking errors approaching simulation results under moderate current conditions. These findings confirm the effectiveness of the proposed framework for underactuated AUV path-following tasks. Full article

In the scallop shucking industry, labor-intensive manual operations and inconsistent product quality remain persistent challenges. To address these issues, this study develops a scallop shucking system based on visual feedback temperature control. To support this system, systematic experiments are conducted to determine the baseline shucking method and its initial parameters. On this basis, the scallop doneness detection model SDD-RT-DETR, serving as the system’s core, integrates an enhanced backbone centered on the self-developed module HierarchicalRepBlock, a frequency-domain self-attention module AIFI-EDFFN, a neck featuring the self-developed EfficientBalanceFusion module as the feature fusion unit and the Converse2DC3 module as the feature extraction unit, and the Wise-DIoU loss function. A scallop doneness image dataset specifically constructed for this task was used to train and validate this model. Experimental results demonstrate that the model achieves 95.5% accuracy, 93.6% recall, and 96.1% mAP50, improving by 4.7%, 3.2%, and 4.2%, respectively, over the baseline model RT-DETR. Additionally, the model’s computational cost was reduced by 18.9%, and its parameters were reduced by 14.1% compared to the baseline model. This model provides real-time, accurate doneness assessment, thereby filling a gap in computer vision for scallop doneness detection. Furthermore, this study integrated this model into a water bath scallop shucking system with feedback temperature control, ensuring that the doneness of scallops after shucking is maintained within an optimal range. Batch tests show that this system achieves a 96.6% shucking rate and an 88.5% properly cooked rate, outperforming the conventional method. This improves the quality of automatically shucked scallops while offering a practical solution for the intelligent upgrading of the seafood processing industry. Full article