Search Results (76)

Understanding how disturbances affect marine foundation species is critical for enhancing the success of coastal ecosystem restoration. Extreme bioturbation by burrowing animals is increasingly impacting coastal vegetated habitats worldwide, with the potential to undermine the persistence and resilience of key foundation species. However, the role of faunal disturbances in modulating restoration outcomes remains poorly understood. Here, we combine field surveys and manipulative field experiments to examine how mud shrimp (Upogebia major) bioturbation impacts vegetation dynamics and restoration outcomes for intertidal seagrass (Zostera japonica). Field surveys revealed pronounced seasonal variation in shrimp bioturbation intensity, with peak burrow densities occurring in fall (up to 400 burrows m⁻²; 289% higher than in spring). The intensified bioturbation was associated with significant declines in seagrass shoot cover, density, and biomass, with negative associations restricted to fall. To test whether seasonally intensified shrimp bioturbation impairs seagrass restoration, we conducted a 24-day field experiment transplanting seagrass patches of varying initial sizes (5–26 cm diameter) into plots representing three levels of shrimp burrow density observed during the fall peak: control (~9 burrows m⁻²), high (~280 burrows m⁻²), and extremely high (~455 burrows m⁻²). Compared to the control, high and extremely high burrow treatments exhibited accelerated patch losses. By day 24, vegetation was virtually eliminated in all shrimp treatments, but the rate of patch loss was significantly lower in larger patches. These results suggest that seasonal intensification of mud shrimp bioturbation has a potential to compromise intertidal seagrass restoration, while increasing planting scale offers a potential mitigation strategy. Restoration interventions should explicitly consider temporal patterns in faunal bioturbation and integration of positive interactions to improve long-term success of vegetation restoration in bioturbator-dominated coastal systems. Full article

With the growing adoption of deep learning in remote sensing, the increasing diversity of models and datasets has made method selection and experimentation more challenging, especially for non-expert users. This study presents a comprehensive evaluation of photovoltaic panel segmentation using a large-scale ultra-high-resolution benchmark of over 25,000 manually annotated unmanned aerial vehicle image patches, systematically quantifying the impact of model and data characteristics. Our results indicate that increasing the spatial diversity of training data has a more substantial impact on training stability and segmentation accuracy than simply adding spectral bands or enlarging the dataset volume. Across all experimental settings, moderate-sized models (DeepLabV3_50, ResUNet50, and SegFormer B4) often provided the best trade-off between segmentation performance and computational efficiency, achieving an average Intersection over Union (IoU) of 0.8966 comparable to 0.8970 of larger models. Moreover, model architecture plays a more critical role than model size; as the ResUNet models consistently achieved higher mean IoU than both DeepLabV3 and SegFormer models, with average improvements of 0.047 and 0.143, respectively. Our findings offer quantitative guidance for balancing architectural choices, model complexity, and dataset design, ultimately promoting more robust and efficient deployment of deep learning models in high-resolution remote sensing applications. Full article

Semantic segmentation of ultra-high-resolution remote sensing (UHR-RS) imagery plays a critical role in land use and land cover analysis, yet it remains computationally intensive due to the enormous input size and high spatial complexity. Existing studies have commonly employed strategies such as patch-wise processing, multi-scale model architectures, lightweight networks, and representation sparsification to reduce resource demands, but they have often struggled to maintain long-range contextual awareness and scalability for inputs of arbitrary size. To address this, we propose RingFormer-Seg, a scalable Vision Transformer framework that enables long-range context learning through multi-device parallelism in UHR-RS image segmentation. RingFormer-Seg decomposes the input into spatial subregions and processes them through a distributed three-stage pipeline. First, the Saliency-Aware Token Filter (STF) selects informative tokens to reduce redundancy. Next, the Efficient Local Context Module (ELCM) enhances intra-region features via memory-efficient attention. Finally, the Cross-Device Context Router (CDCR) exchanges token-level information across devices to capture global dependencies. Fine-grained detail is preserved through the residual integration of unselected tokens, and a hierarchical decoder generates high-resolution segmentation outputs. We conducted extensive experiments on three benchmarks covering UHR-RS images from 2048 × 2048 to 8192 × 8192 pixels. Results show that our framework achieves top segmentation accuracy while significantly improving computational efficiency across the DeepGlobe, Wuhan, and Guangdong datasets. RingFormer-Seg offers a versatile solution for UHR-RS image segmentation and demonstrates potential for practical deployment in nationwide land cover mapping, supporting informed decision-making in land resource management, environmental policy planning, and sustainable development. Full article

Matching drone images to satellite reference images is a critical step for achieving UAV self-localization. Existing drone visual localization datasets mainly focus on target localization, where each drone image is paired with a corresponding satellite image slice, typically with identical coverage. However, this one-to-one approach does not reflect real-world UAV self-localization needs as it cannot guarantee exact matches between drone images and satellite tiles nor reliably identify the correct satellite slice. To bridge this gap, we propose a one-to-many matching method between drone images and satellite reference tiles. First, we enhance the UAV-VisLoc dataset, making it the first in the field tailored for one-to-many imperfect matching in UAV self-localization. Second, we introduce a novel loss function, Incomp-NPair Loss, which better reflects real-world imperfect matching scenarios than traditional methods. Finally, to address challenges such as limited dataset size, training instability, and large-scale differences between drone images and satellite tiles, we adopt a Vision Transformer (ViT) baseline and integrate CNN-extracted features into its patch embedding layer. Full article

Limited depth of field in modern optical imaging systems often results in partially focused images. Multi-focus image fusion (MFF) addresses this by synthesizing an all-in-focus image from multiple source images captured at different focal planes. While deep learning-based MFF methods have shown promising results, existing approaches face significant challenges. Convolutional Neural Networks (CNNs) often struggle to capture long-range dependencies effectively, while Transformer and Mamba-based architectures, despite their strengths, suffer from high computational costs and rigid input size constraints, frequently necessitating patch-wise fusion during inference—a compromise that undermines the realization of a true global receptive field. To overcome these limitations, we propose MLP-MFF, a novel lightweight, end-to-end MFF network built upon the Pyramid Fusion Multi-Layer Perceptron (PFMLP) architecture. MLP-MFF is specifically designed to handle flexible input scales, efficiently learn multi-scale feature representations, and capture critical long-range dependencies. Furthermore, we introduce a Dual-Path Adaptive Multi-scale Feature-Fusion Module based on Hybrid Attention (DAMFFM-HA), which adaptively integrates hybrid attention mechanisms and allocates weights to optimally fuse multi-scale features, thereby significantly enhancing fusion performance. Extensive experiments on public multi-focus image datasets demonstrate that our proposed MLP-MFF achieves competitive, and often superior, fusion quality compared to current state-of-the-art MFF methods, all while maintaining a lightweight and efficient architecture. Full article

The axle-box bearing is a critical load-bearing component in high-speed trains and is prone to failure under long-term heavy-duty operation, affecting both operational efficiency and safety. Current deep-learning-based fault diagnosis methods face two key challenges: difficulty in capturing temporal features across multiple scales simultaneously, and limited capability in modeling local sequential patterns. To address these issues, we propose P2IFormer, a fault diagnosis model based on multi-granularity patch-to-image embedding. The raw vibration sequence is divided into equal-length patch sequences under multiple granularities, each defined by a fixed window size. Each patch is then transformed into a Gramian Angular Field (GAF) image to extract spatial features and generate granularity-specific embedding. A multi-granularity self-attention mechanism is used to model both intra- and inter-granularity dependencies. The resulting multi-granularity features are fused and fed into a softmax classifier for final fault prediction. Experiments conducted under four constant-speed conditions and one variable-speed condition demonstrate that P2IFormer achieves over 99.5% accuracy across all scenarios, significantly outperforming existing CNN- and Transformer-based methods. Full article

Breast cancer detection through non-invasive and accurate techniques remains a critical challenge in medical diagnostics. This study introduces a deep learning-based framework that leverages a microwave radar system equipped with an arc-shaped array of six antennas to estimate key tumor parameters, including position, size, and depth. This research begins with the evolutionary design of an ultra-wideband octagram ring patch antenna optimized for enhanced tumor detection sensitivity in directional near-field coupling scenarios. The antenna is fabricated and experimentally evaluated, with its performance validated through S-parameter measurements, far-field radiation characterization, and efficiency analysis to ensure effective signal propagation and interaction with breast tissue. Specific Absorption Rate (SAR) distributions within breast tissues are comprehensively assessed, and power adjustment strategies are implemented to comply with electromagnetic exposure safety limits. The dataset for the deep learning model comprises simulated self and mutual S-parameters capturing tumor-induced variations over a broad frequency spectrum. A core innovation of this work is the development of the Attention-Based Feature Separation (ABFS) model, which dynamically identifies optimal frequency sub-bands and disentangles discriminative features tailored to each tumor parameter. A multi-branch neural network processes these features to achieve precise tumor localization and size estimation. Compared to conventional attention mechanisms, the proposed ABFS architecture demonstrates superior prediction accuracy and interpretability. The proposed approach achieves high estimation accuracy and computational efficiency in simulation studies, underscoring the promise of integrating deep learning with conformal microwave imaging for safe, effective, and non-invasive breast cancer detection. Full article

Microcalcification clusters (MCCs) are key indicators of breast cancer, with studies showing that approximately 50% of mammograms with MCCs confirm a cancer diagnosis. Early detection is critical, as it ensures a five-year survival rate of up to 99%. However, MCC detection remains challenging due to their features, such as small size, texture, shape, and impalpability. Convolutional neural networks (CNNs) offer a solution for MCC detection. Nevertheless, CNNs are typically trained on single-resolution images, limiting their generalizability across different image resolutions. We propose a CNN trained on digital mammograms with three common resolutions: 50, 70, and 100 $\mathsf{\mu }$m. The architecture processes individual 1 cm² patches extracted from the mammograms as input samples and includes a MobileNetV2 backbone, followed by a flattening layer, a dense layer, and a sigmoid activation function. This architecture was trained to detect MCCs using patches extracted from the INbreast database, which has a resolution of 70 $\mathsf{\mu }$m, and achieved an accuracy of 99.84%. We applied transfer learning (TL) and trained on 50, 70, and 100 $\mathsf{\mu }$m resolution patches from the MEXBreast database, achieving accuracies of 98.32%, 99.27%, and 89.17%, respectively. For comparison purposes, models trained from scratch, without leveraging knowledge from the pretrained model, achieved 96.07%, 99.20%, and 83.59% accuracy for 50, 70, and 100 $\mathsf{\mu }$m, respectively. Results demonstrate that TL improves MCC detection across resolutions by reusing pretrained knowledge. Full article

Reservoir-induced landslides threaten infrastructures and downstream communities, making continuous deformation monitoring vital. Time-series InSAR, notably the SBAS algorithm, provides high-precision surface-displacement mapping but suffers from voids due to layover/shadow effects and temporal decorrelation. Existing deep-learning approaches often operate on fixed-size patches or ignore irregular spatio-temporal dependencies, limiting their ability to recover missing pixels. With this objective, a hybrid spatio-temporal Graph Attention (ST-GAT) framework was developed and trained on SBAS-InSAR values using 24 influential features. A unified spatio-temporal graph is constructed, where each node represents a pixel at a specific acquisition time. The nodes are connected via inverse distance spatial edges to their K-nearest neighbors, and they have bidirectional temporal edges to themselves in adjacent acquisitions. The two spatial GAT layers capture terrain-driven influences, while the two temporal GAT layers model annual deformation trends. A compact MLP with per-map bias converts the fused node embeddings into normalized LOS estimates. The SBAS-InSAR results reveal LOS deformation, with 48% of missing pixels and 20% located near the Dasu dam. ST D-GAT reconstructed fully continuous spatio-temporal displacement fields, filling voids at critical sites. The model was validated and achieved an overall R² (0.907), ρ (0.947), per-map R² ≥ 0.807 with RMSE ≤ 9.99, and a ROC-AUC of 0.91. It also outperformed the six compared baseline models (IDW, KNN, RF, XGBoost, MLP, simple-NN) in both RMSE and R². By combining observed LOS values with 24 covariates in the proposed model, it delivers physically consistent gap-filling and enables continuous, high-resolution landslide monitoring in radar-challenged mountainous terrain. Full article

Alzheimer’s disease (AD) involves the accumulation of amyloid-β (Aβ) plaques, whose quantification plays a central role in understanding disease progression. Automated segmentation of Aβ deposits in histopathological micrographs enables large-scale analyses but is hindered by the high cost of detailed pixel-level annotations. Weakly supervised learning offers a promising alternative by leveraging coarse or indirect labels to reduce the annotation burden. We evaluated a weakly supervised approach to segment and analyze thioflavin-S-positive parenchymal amyloid pathology in AD and age-matched brains. Our pipeline integrates three key components, each designed to operate under weak supervision. First, robust preprocessing (including retrospective multi-image illumination correction and gradient-based background estimation) was applied to enhance image fidelity and support training, as models rely more on image features. Second, class activation maps (CAMs), generated by a compact deep classifier SqueezeNet, were used to identify, and coarsely localize amyloid-rich parenchymal regions from patch-wise image labels, serving as spatial priors for subsequent refinement without requiring dense pixel-level annotations. Third, a patch-based convolutional neural network, U-Net, was trained on synthetic data generated from micrographs based on CAM-derived pseudo-labels via an extensive object-level augmentation strategy, enabling refined whole-image semantic segmentation and generalization across diverse spatial configurations. To ensure robustness and unbiased evaluation, we assessed the segmentation performance of the entire framework using patient-wise group k-fold cross-validation, explicitly modeling generalization across unseen individuals, critical in clinical scenarios. Despite relying on weak labels, the integrated pipeline achieved strong segmentation performance with an average Dice similarity coefficient (≈0.763) and Jaccard index (≈0.639), widely accepted metrics for assessing segmentation quality in medical image analysis. The resulting segmentations were also visually coherent, demonstrating that weakly supervised segmentation is a viable alternative in histopathology, where acquiring dense annotations is prohibitively labor-intensive and time-consuming. Subsequent morphometric analyses on automatically segmented Aβ deposits revealed size-, structural complexity-, and global geometry-related differences across brain regions and cognitive status. These findings confirm that deposit architecture exhibits region-specific patterns and reflects underlying neurodegenerative processes, thereby highlighting the biological relevance and practical applicability of the proposed image-processing pipeline for morphometric analysis. Full article

The precise identification of priority areas for conservation based on connectivity can significantly enhance protection efficacy and mitigate biodiversity loss in fragmented landscapes. Priority area selection efforts are typically conducted in landscapes with a limited number of patches or simplified to focus on large patches, while landscapes with numerous patches are rarely explored. In this paper, we used a forest in Kanas, Xinjiang, China, as a case study to explore priority patches for conservation according to their contribution to maintaining overall landscape connectivity, as well as to assess how structural factors influence patch importance in connectivity, based on graph theory. We found that the rank of patches varied with patch importance indices (which can be used to calculate the contribution of individual patches to maintaining overall landscape). Dispersal distances were selected, as they placed different emphasis on the size and topological location of patches, and different types of links (binary or probabilistic connection) were used. One critical and seven important connected patches were identified as priority patches for conservation after taking multiple connectivity indices and dispersal distances into comprehensive consideration. In addition, neighboring patch density was the dominant factor that influenced patch importance for species with 50 and 100 m dispersal distances, while patch size contributed most for species with 200 m and longer dispersal distances; therefore, we suggested that neighboring patch density and patch size could be used to support efforts to identify priority patches. Overall, our results provide a unique perspective and a more simplified process for the selection of priority protected sites in patch-rich landscapes, allowing us to highlight which action is suitable for optimizing landscape connectivity and biodiversity conservation. Full article

Mangroves form the dominant coastal plant community of low-energy tropical intertidal habitats and provide critical ecosystem services to humans and the environment. However, more frequent and increasingly powerful hurricanes and storm surges are creating additional pressure on the natural resilience of these threatened coastal ecosystems. Advances in remote sensing techniques and approaches are critical to providing robust quantitative monitoring of post-storm mangrove forest recovery to better prioritize the often-limited resources available for the restoration of these storm-damaged habitats. Here, we build on previously utilized spatial and temporal ranges of European Space Agency (ESA) Sentinel satellite imagery to monitor and map the recovery of the mangrove forests of the British Virgin Islands (BVI) since the occurrence of back-to-back category 5 hurricanes, Irma and Maria, on September 6 and 19 of 2017, respectively. Pre- to post-storm changes in coastal mangrove forest health were assessed annually using the normalized difference vegetation index (NDVI) and moisture stress index (MSI) from 2016 to 2023 using Google Earth Engine. Results reveal a steady trajectory towards forest health recovery on many of the Territory’s islands since the storms’ impacts in 2017. However, some mangrove patches are slower to recover, such as those on the islands of Virgin Gorda and Jost Van Dyke, and, in some cases, have shown a continued decline (e.g., Prickly Pear Island). Our work also uses a linear ANCOVA model to assess a variety of geospatial, environmental, and anthropogenic drivers for mangrove recovery as a function of NDVI pre-storm and post-storm conditions. The model suggests that roughly 58% of the variability in the 7-year difference (2016 to 2023) in NDVI may be related by a positive linear relationship with the variable of population within 0.5 km and a negative linear relationship with the variables of northwest aspect vs. southwest aspect, island size, temperature, and slope. Full article

Pedestrian detection systems are widely used in safety-critical domains such as autonomous driving, where deep neural networks accurately perceive individuals and distinguish them from other objects. However, their vulnerability to backdoor attacks remains understudied. Existing backdoor attacks, relying on unnatural digital perturbations or explicit patches, are difficult to deploy stealthily in the physical world. In this paper, we propose a novel backdoor attack method that leverages real-world occlusions (e.g., backpacks) as natural triggers for the first time. We design a dynamically optimized heuristic-based strategy to adaptively adjust the trigger’s position and size for diverse occlusion scenarios, and develop three model-independent trigger embedding mechanisms for attack implementation. We conduct extensive experiments on two different pedestrian detection models using publicly available datasets. The results demonstrate that while maintaining baseline performance, the backdoored models achieve average attack success rates of 75.1% on KITTI and 97.1% on CityPersons datasets, respectively. Physical tests verify that pedestrians wearing backpack triggers could successfully evade detection under varying shooting distances of iPhone cameras, though the attack failed when pedestrians rotated by 90°, confirming the practical feasibility of our method. Through ablation studies, we further investigate the impact of key parameters such as trigger patterns and poisoning rates on attack effectiveness. Finally, we evaluate the defense resistance capability of our proposed method. This study reveals that common occlusion phenomena can serve as backdoor carriers, providing critical insights for designing physically robust pedestrian detection systems. Full article

Diffusiophoresis of a weakly charged dielectric droplet in a cylindrical pore is investigated theoretically in this study. The governing fundamental electrokinetic equations are solved with a patched pseudo-spectral method based on Chebyshev polynomials, coupled with a geometric mapping scheme to take care of the irregular solution domain. The impact of the boundary confinement effect upon the droplet motion is explored in detail, which is most profound in narrow channels. We found, among other things, that the droplet moving direction may reverse with varying channel widths. Enhanced motion-inducing double-layer polarization due to the presence of a nearby channel wall is found to be responsible for it. In particular, an interesting and seemingly peculiar phenomenon referred to as the “solidification phenomenon” is observed here at some specific critical droplet sizes or electrolyte strengths in narrow channels, under which all the droplets move at identical speeds regardless of their viscosities. They move like a rigid particle without the surface spinning motions and the induced interior recirculating vortex flows. As the corresponding shear rate is zero at this point, the droplet is resilient to undesirable exterior shear stresses tending to damage the droplet in motion. This provides a helpful guideline in the fabrication of liposomes in drug delivery in terms of the optimal liposome size, as well as in the microfluidic and nanofluidic manipulations of cells, among other potential practical applications. The effects of other parameters of electrokinetic interest are also examined. Full article

South and Southeast Asia are considered biodiversity hotspots, yet they face escalating threats from deforestation and climate change. This study evaluates the suitable habitat extent of two sympatric flying squirrels, Petaurista elegans and Hylopetes phayrei, using ensemble distribution models based on the climate-only model (COM) and habitat–climate model (HCM) approaches. The results indicated severe habitat loss, with suitable areas comprising only 1.56–1.66% (P. elegans) and 0.22–2.47% (H. phayrei) of their estimated extent of occurrence. Within IUCN-defined ranges, the suitability for P. elegans was 28.25% and 30.04%, while H. phayrei showed 2.86% and 32.39% in terms of the HCM and COM, respectively. The analysis further revealed habitat fragmentation, reduced patch size, and edge complexity, with future scenarios predicting increased isolation. These results highlight the urgent necessity for region-specific conservation strategies focusing on habitat recovery, connectivity, and transboundary cooperation. The recommended actions include genetic studies, corridor analysis, and field validation. This research provides critical baseline data to inform integrated, multi-stakeholder conservation planning across South and Southeast Asia for the long-term persistence of these vulnerable flying squirrel species. Full article