Search Results (106)

Detecting small targets in Unmanned Aerial Vehicle (UAV) imagery is challenging due to low pixel coverage, complex backgrounds, and information loss during downsampling. Existing detectors lack explicit mechanisms for enhancing weak target signals. We propose GHF-DETR, a Transformer-based detector featuring three collaboratively designed modules. First, a Heterogeneous Multi-Path Convolutional Network (HMC) backbone uses partial convolution and gated linear units to reduce computational redundancy while maintaining discrimination of small-object features. Second, a Dynamic Multi-Scale Focusing (DMSF) module integrates learned offset alignment with multi-kernel depthwise convolutions for cross-scale feature fusion. Third, a High-Frequency Selective Preservation (HSP) downsampling module combines space-to-depth convolution with 2D Discrete Wavelet Transform (DWT) to compensate for information loss in both spatial and frequency domains. On VisDrone2019, GHF-DETR achieves 33.1% mAP@0.5 and 18.6% mAP@0.5:0.95 with 15.4 GFLOPs and 7.59 M parameters, improving over the DFINE-n baseline by 5.4% and 3.1%, respectively, with AP_S reaching 10.1%. Generalization is validated on NWPU VHR-10. These results demonstrate that GHF-DETR achieves a favorable accuracy–efficiency balance for efficient UAV small-object detection. Full article

Radio frequency (RF) energy harvesting offers a potential power source for low-power Internet of Things and wireless sensing nodes, but compact rectifiers must balance impedance matching, multiband response, and load-driving capability. This work presents a compact SMS7621 Schottky-diode RF rectifier for RF-powered wireless sensing applications. An 11-segment microstrip distributed-parameter collaborative optimization strategy is used to tune impedance transformation in a 3.48 cm × 1.98 cm single-layer layout while compensating for diode nonlinear impedance variation and package parasitics. Simulations show more than 40% RF-to-DC conversion efficiency from 1.90 to 2.35 GHz, with additional efficiency peaks of 40.55% at 4.45 GHz and 38.45% at 7.15 GHz. Measurements verify the 2.45 GHz output performance under controlled high-input-power excitation: with a 300 Ω load and 25 dBm input, the rectifier delivers a maximum DC voltage of 5.42 V. At 15 dBm input, the measured peak efficiency reaches 46.05% at 2 GHz and remains 35.69% at 4 GHz. These results indicate a compact rectifier front end with multiband harvesting potential and 5 V-class load-driving capability under dedicated RF powering conditions. Full article

In distributed Geographic Information Systems (GIS), preserving topological validity without sacrificing real-time interactivity under high-frequency concurrent editing of spatial polygons remains a persistent challenge. Recent distance-based heuristic methods suffer from scale-dependent bottlenecks and unreliable topology preservation, while more robust application-layer caching mechanisms still incur severe queuing latency under intense concurrency. To overcome these limitations, we propose Geo-CRDT, a geometry-aware distributed data structure that integrates spatial constraints directly into its underlying architecture. By dynamically isolating concurrent spatial entanglements into a strictly bounded local scope $\mathcal{S}$, the system deterministically resolves complex 2D conflicts via scalar projection, repairing the local topology in $O\left(\right|\mathcal{S}\left|\right)$ time. Rigorous simulations and a 15-participant real-world case study validate that Geo-CRDT sustains low-latency responsiveness and structural reliability under extreme concurrency, offering a robust foundation for large-scale crowdsourced spatial collaboration. Full article

To address the transient frequency limit violations triggered by the low-inertia characteristics of temporary islanded microgrids formed under extreme disasters, this paper proposes a multi-source collaborative two-stage robust optimization day-ahead scheduling model considering dynamic frequency constraints. Firstly, a collaborative architecture encompassing emergency power vehicles, grid-forming energy storage systems, and flexible loads is constructed. Through collaborative scheduling in the day-ahead pre-scheduling and real-time re-scheduling stages, this architecture effectively avoids the exorbitant costs of physical load shedding under extreme conditions. Secondly, to overcome the limitations of traditional robust box uncertainty sets—which ignore temporal correlations, tend to cause non-physical high-frequency oscillations, and hinder algorithm convergence—a time-correlated uncertainty set based on state-transition auxiliary variables is designed to accurately capture the continuous evolution characteristics of meteorological disturbances. The column-and-constraint generation algorithm is utilized for the solution methodology, combined with the big-M method to transform the subproblem containing bilinear terms into a mixed-integer linear programming model for efficient solving. Simulation results on a modified 33-node test system demonstrate that the proposed model effectively filters out high-frequency oscillation trajectories and significantly improves computational efficiency. Under the worst-case temporal disturbances, the transient frequency drop and the rate of change in frequency are strictly controlled within safe thresholds. Compared to deterministic scheduling and traditional box-based robust models, the proposed scheme effectively balances system security and economic efficiency, demonstrating exceptional system resilience and defense capabilities against varying prediction errors. Full article

To address the issues of insufficient single-modality robustness and limited multi-scale object detection accuracy in remote sensing image detection (RSID) in complex environments, this paper proposes a multimodal RSID network named VI-MSFFN. The model adopts a symmetric parallel dual-branch architecture to achieve independent extraction and collaborative modeling of visible and infrared modal features. A cross-modal multi-scale sparse cross-attention fusion module is proposed and applied to the P4 and P5 feature layers, and a high-low-level feature collaborative cross-modal fusion strategy was constructed to achieve efficient and robust cross-modal feature fusion while enhancing multi-scale object modeling capability and suppressing feature redundancy and noise. Additionally, a progressive feature interaction and fusion architecture was designed to combine spatial and frequency domain information to strengthen deep object representation. The experimental results on the VEDAI and Drone Vehicle datasets demonstrate that VI-MSFFN achieves state-of-the-art (SOTA) performance in detection accuracy, robustness, and generalization ability. The proposed method effectively solves the detection challenges of RSID and has significant application value in the field of multi-modal RSID. Full article

Small object detection in UAV aerial imagery poses significant challenges due to sparse pixel representation and ambiguous object boundaries. Through systematic analysis, we identify three critical degradation stages during forward propagation in deep detection networks: edge attenuation during feature extraction, semantic conflict during feature fusion, and detail loss during feature reconstruction. Existing methods address these stages in isolation or implicitly, lacking collaborative and stage-aware repair strategies. To address this issue, we propose EAGLE-DET, a novel detection framework based on sparse multi-scale attention and refined transformation. Specifically, the framework comprises three core modules: (1) the Cross-stage Multi-resolution Edge Enhancement Network (CMENet), which preserves small object edge representations via adaptive high-low frequency decomposition; (2) the Attention-guided Multi-scale Feature Fusion Network (AMFFN), which resolves cross-scale semantic conflicts through pyramidal sparse attention and multi-scale spatial decoupling; (3) the Enhanced Upsampling with Channel Bridging and Spatial Coordination module (EUCBSC), which recovers spatial detail fidelity via bidirectional channel shift mixing. Extensive experiments on three benchmark datasets—VisDrone-2019, UAVDT, and DOTA1.0—demonstrate the effectiveness of EAGLE-DET, which achieves improvements of 4.5% AP50 and 2.9% AP50:95 on VisDrone-2019 over the baseline, while maintaining inference at 71.7 FPS, achieving an optimal accuracy–efficiency trade-off. Full article

Short-term wind speed forecasting plays a critical role in enabling the reliable integration of renewable energy and supporting the sustainable operation of power systems. However, traditional dual-frequency decomposition methods oversimplify wind speed dynamics by separating them into only high-frequency disturbances and low-frequency trends, making it difficult to capture intermediate-frequency transitional dynamics. Additionally, single models struggle to adapt to multi-scale temporal features, limiting forecasting performance. To address these issues, this paper proposes a tri-band frequency-aware heterogeneous expert collaboration framework. First, Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (CEEMDAN) is employed for signal denoising, followed by Particle Swarm Optimization-Time Varying Filtering-based Empirical Mode Decomposition (PSO-TVF-EMD) for multi-scale signal disentanglement. Then, Permutation Entropy (PE) is used to construct a tri-band structure consisting of high-, intermediate-, and low-frequency components. A frequency-aware expert routing mechanism assigns Bayesian Optimization Long Short-Term Memory (BO-LSTM), an improved Markov model, and Auto-Regressive Integrated Moving Average (ARIMA) to the corresponding frequency bands. Finally, a reliability-aware cooperative aggregation strategy integrates predictions from multiple experts. Experimental results show that representative baseline models, including BO-LSTM, Markov, ARIMA, Gated Recurrent Unit (GRU) and Convolutional Neural Network Long Short-Term Memory (CNN-LSTM), achieve MAE values ranging from 0.308 to 0.429, while the proposed framework reduces the Mean Absolute Error (MAE) to 0.193 and Root Mean Square Error (RMSE) to 0.274, with a Mean Absolute Percentage Error (MAPE) of 7.35% and R² of 0.927. Compared with the dual-frequency decomposition scheme (MAE = 0.266), the proposed tri-band framework achieves an average improvement of approximately 28.1%. The results suggest that explicitly modeling intermediate-frequency dynamics and aligning model inductive biases with multi-scale signal characteristics can effectively enhance short-term wind speed forecasting performance. Full article

In practical imaging applications, low-light and overexposure are two common types of image degradation problems with inherent conflicts, and existing methods struggle to achieve accurate restoration of both degradations within a unified framework. To address this challenge, this paper proposes DFE-Net based on explicit frequency decoupling. The network adopts a symmetric U-Net architecture and embeds discrete wavelet transform (DWT) and inverse discrete wavelet transform (IWT) to construct an explicit dual-frequency processing mechanism, which optimizes the low-frequency information carrying global illumination and the high-frequency information containing detailed textures, respectively. In the encoder, DWT decouples features into low-frequency and high-frequency sub-bands and feeds them into dedicated enhancement modules. The low-frequency enhancement block integrates SS2D and a gated convolutional feed-forward network to efficiently model global contextual dependencies with linear complexity and accurately restore image illumination and contrast; the high-frequency enhancement block adopts CMT attention combined with a matching convolutional feed-forward network, enabling the detail restoration process to be guided by the optimized low-frequency information and ensuring the collaborative optimization of global structure and local textures. The decoder completes the reconstruction and fusion of the processed sub-bands through IWT. The quantitative and qualitative experimental results on the MSEC, SICE, and LOLv1 datasets demonstrate that DFE-Net achieves or surpasses existing state-of-the-art methods in various metrics while maintaining low model complexity. Full article

Traditional high-precision canopy segmentation heavily relies on tedious pixel-level manual annotation, while general-purpose zero-shot visual detection algorithms are prone to boundary adhesion and excessive computational load in dense forest areas. To address this, this study proposes a human–machine collaborative, efficient canopy segmentation and canopy cover inversion paradigm, combining the zero-shot pre-annotation capabilities of text-driven object detection with the high-precision segmentation advantages of the lightweight proprietary network LGBU-Net. In the offline annotation stage, this method automatically locates candidate canopy regions using Grounding DINO combined with text prompts and generates initial pixel-level masks using SAM. A high-quality training set is then constructed through minimal manual correction, significantly reducing the cost of traditional fully manual annotation. Subsequently, an improved LGBU-Net designed for complex forest conditions is used for supervised learning. In the feature extraction stage, a lightweight phantom-coordinate attention module (LG-CAM) is introduced to enhance the network’s focus on the geometric center of the tree canopy and suppress semantic interference caused by the forest background, light spots, and shadows. In the decoding stage, a boundary difference fusion module (BDF-Block) is deployed to alleviate the problem of adjacent tree canopy boundaries adhering by utilizing high-frequency gradient information from the underlying layers of UAV imagery. Combined with a boundary-aware hybrid loss function, the clarity of individual tree boundaries is further improved in the gradient domain. Experiments based on UAV imagery of high-density mixed and coniferous forests in Baishan, Jilin Province, show that, with low manual annotation costs, LGBU-Net achieves a canopy segmentation IoU of 90.45% and an individual tree separation F1 score of 89.35%, significantly outperforming general visual algorithms with zero-shot direct inference, and with only 4.85 M model parameters. Furthermore, the segmentation results are used for plot-level canopy vertical cover (CC) inversion, and the estimated values are highly consistent with ground-based measurements. This research provides a high-precision, low-annotation-cost technical solution with good edge deployment potential for large-scale forest resource surveys and forest understory light environment assessment. Full article

Existing deep-learning-based motor imagery (MI) electroencephalogram (EEG) decoding methods face challenges in generalizing across sessions and providing channel-level physiological interpretability. These limitations hinder the practical application of MI-EEG systems. Accordingly, DO-PI-EATCNet (Dream-Optimization-Enhanced, Physics-Inspired, Efficient-Attention Temporal Channel Network) is proposed to improve generalization and interpretability in MI-EEG classification. Unlike models that simply combine multiple components, DO-PI-EATCNet assigns distinct roles to feature representation, temporal channel modeling, temporal regularization, and channel compactness. Latent-Projected Attention (LPA) enhances spatiotemporal discriminability by aligning attention in a low-dimensional latent space, and Temporal Channel Cascaded Collaborative Attention (TCCA) refines dependencies between time and channels. Fractional-Order Difference Temporal Consistency Loss (FD-TCL) is introduced as a neurodynamics-inspired temporal regularizer to reduce high-frequency fluctuations in prediction sequences and improve within-subject cross-session prediction stability. The Multi-Population Dream Optimization Algorithm (MPDOA) is used for channel selection to obtain a compact EEG channel subset and reduce computational load, although it introduces a slight accuracy decrease compared with the uncompressed full model. Under a within-subject cross-session protocol on the BCI Competition IV-2a four-class MI dataset, the final compact model achieves an average accuracy of 84.4% and Cohen’s κ of 0.790, outperforming the reimplemented baselines. Compared with the uncompressed LPA-TCCA-FD-TCL variant, MPDOA slightly decreases accuracy from 84.9% to 84.4%, but reduces EEG channels from 22 to about 15 and decreases MACs by 27%. Scalp topographies and selected-channel visualizations provide qualitative support for channel-level anatomical plausibility, as the selected electrodes are mainly located over expected sensorimotor-related regions, while t-SNE offers a descriptive visualization of the learned feature distributions. Full article

Early-Warning Radar (EWR) is an advanced detection system capable of monitoring aerial targets over long distances with high precision, providing critical information support for defense security. However, EWR faces challenges such as a limited number of pulses, low coherent integration gain, small target Radar Cross Section (RCS), and complex clutter and electromagnetic interference environments. Conventional Constant False Alarm Rate (CFAR) detection algorithms struggle to effectively detect weak targets while maintaining an acceptable false alarm rate. To address these issues, this paper introduces a deep learning approach. A high target-clutter/interference/noise discriminative feature spectrum is obtained through phase difference transformation, upon which a dual-branch collaborative architecture network is constructed. In this architecture, the main network focuses on extracting spatiotemporal amplitude–phase characteristics, while the auxiliary branch implicitly mines the target’s physical boundary features from frequency-domain echoes. Through a self-attention mechanism, the features from both branches are semantically aligned and fused. This method significantly enhances the weak target detection capability of EWR under the constraint of a controlled false alarm rate. Test results show that under the false alarm rate ranging from ${10}^{-3}$ to ${10}^{-4}$, the SNR gain of the proposed algorithm is about 2∼5 dB, which is equivalent to increasing the radar detection range by 10∼30%. Full article

High-fidelity infrared remote sensing imagery serves as a critical foundation for the development of technologies such as infrared scene simulation and long-range imaging detection. Addressing the core limitations of two categories of methods: traditional physical modeling methods—low fidelity and efficiency—and deep learning-based generation methods with insufficient interpretability and weak generalization capabilities, we propose a visible-to-infrared (Vis-to-IR) remote sensing image generation method based on the multi-dimensional features of scene elements and mechanism-data dual-driven strategy (MDD-VIR) in this paper. First, a scene element multi-dimensional feature extractor (SEMFE) is designed by analyzing and reconstructing limited datasets, bridging physical mechanisms and intelligent learning. From a game-theoretic perspective, we present a Unet3+-based frequency-domain adaptive spatial channel reconstruction convolution module (FASCRC_Unet3+) and a feature fusion discrimination method based on proactive material weighting (FFD_PMW) to enhance the model’s ability to learn and transform high-value regional and multi-scale features. Furthermore, a collaborative optimization loss function (Loss_CO) is designed to integrate dual-driven paradigm advantages to facilitate efficient iteration. Experiments show that the average SSIM of MDD-VIR simulated images reached 91.07%. Innovatively fusing physical algorithms with intelligent models, this approach enables the Vis-to-IR remote sensing image generation model to achieve the multiple objectives of robust physical consistency, high fidelity, and high efficiency. Full article

The increasing integration of wind power reduces system inertia and weakens the frequency regulation capability of power systems. To address the problems of unclear task allocation, repeated compensation, and insufficient consideration of wind turbine energy limits in wind–thermal coordinated frequency regulation, this paper proposes a multi-band collaborative frequency regulation strategy with dynamic energy-constrained control. The proposed method decomposes the frequency deviation into high- and low-frequency components, enabling wind turbines to provide fast support and thermal units to undertake sustained regulation. A residual-based cross-band feedback mechanism is introduced to reduce repeated compensation, while dynamic segmented thresholds and releasable power constraints are used to adaptively adjust wind turbine support according to available energy. Simulations on a modified IEEE 3-machine 9-bus system show that, compared with the conventional strategy, the proposed method reduces the secondary frequency drop from 0.028 Hz to 0.015 Hz and shortens the recovery time from 20.7 s to 16.6 s. The results indicate that the proposed method can provide a practical reference for coordinated primary frequency regulation in wind–thermal power systems. Full article

Maritime cross-domain unmanned system-of-systems (MCUSoS), featuring multi-domain collaboration, wide-area coverage, and flexible deployment, plays a vital role in missions such as maritime search and rescue, marine environmental monitoring, and terrain reconnaissance. MCUSoS enables collaborative detection by coordinating heterogeneous unmanned clusters across the aerial, surface, and underwater domains. However, this capability is vulnerable to degradation under cross-domain heterogeneity, communication constraints, and external disturbances such as node failures, link disruptions and malicious interference. To address these challenges, this paper proposes an integrated framework for collaborative detection capability evaluation and resilience enhancement of MCUSoS in multi-disturbance environments. Firstly, a system-of-systems architecture is established by incorporating formation detection modes and multi-level collaborative relationships to characterize its collaborative detection capabilities. Second, a capability evaluation model is developed from the capabilities of collaboration and detection. Based on this, a multi-stage resilience evaluation mechanism is proposed to quantify MCUSoS resilience under three disturbance modes. Additionally, a resilience enhancement strategy combining internal reconfiguration with the external deployment of supplementary detection nodes is designed to recover MCUSoS performance in multi-disturbance environments. Finally, a case study involving 12 clusters of MCUSoS is conducted to validate the effectiveness of the proposed methods. The results demonstrate that the proposed resilience enhancement strategy achieves a recovery rate of up to 74% in the disintegration circle attack scenario and consistently improves the resilience of the MCUSoS under targeted attacks, with the resilience value under low-frequency attacks being 148% higher than that under high-frequency attacks. These findings provide a quantitative basis for resilience evaluation and enhancement in dynamic scenarios. Full article

Background/Objectives: The accumulation of unused and expired pharmaceuticals in households is a growing public health concern with implications for patient safety, rational drug use, and environmental health. However, systematic risk characterization integrating clinical and environmental perspectives at the community level remains limited, particularly in low- and middle-income settings. This study aimed to develop and apply a composite risk index, grounded in an eco-pharmacovigilance framework, for the assessment of health risks associated with accumulated household pharmaceuticals in southeastern Mexico. Methods: A cross-sectional study was conducted in 526 randomly selected households using stratified sampling. Guided in-home medication inventories were performed with participant collaboration, and pharmaceuticals were classified according to the Anatomical Therapeutic Chemical (ATC) system. A composite risk index (CRI = Fr × PR) was developed within an eco-pharmacovigilance framework. The frequency of accumulation (Fr) for each therapeutic group was multiplied by a potential risk score (PR) derived through a structured multidisciplinary expert consensus process integrating clinical toxicity, environmental persistence, and antimicrobial resistance potential. Results: A total of 2184 pharmaceutical units were recorded during the household inventories, of which 28.7% were expired. Expired medications were primarily retained rather than actively used, representing a latent risk for inappropriate self-medication and accidental exposure. The therapeutic groups with the highest CRI values were antihypertensives (CRI = 42.3), antidiabetics (CRI = 37.8), and antibiotics (CRI = 31.5), indicating a relatively higher contribution within the composite risk index framework to overall household pharmaceutical risk. These findings highlight priority therapeutic groups driven by the combined effect of high accumulation frequency, distinct accumulation patterns, and intrinsic hazard. Conclusions: Household pharmaceutical accumulation can be characterized using a composite, eco-pharmacovigilance-based approach that integrates exposure and hazard dimensions. The proposed framework functions as a prioritization tool rather than a precise quantitative measure, enabling the identification of therapeutic groups requiring targeted intervention. Findings should be interpreted as indicative of relative risk patterns rather than precise estimates, given the exploratory design and guided data collection approach. The proposed framework provides a practical tool for prioritizing interventions aimed at improving rational drug use, reducing accumulation, and mitigating environmental impact. Further validation in diverse settings is warranted to strengthen its applicability. Full article