Search Results (1,240)

Accurate detection of surface defects in wind turbine blades is critical for condition monitoring and preventive maintenance of wind energy systems. Defects such as cracks, burns, deformation, and peeling are characterized by small dimensions, irregular morphologies, and low contrast, limiting the effectiveness of conventional feature extraction methods. Although YOLOv13 enhances high-order feature correlation and information flow, its fixed-grid spatial sampling and content-agnostic upsampling operations remain limited in adapting to irregular defect geometries and preserving fine-grained boundary information. This study proposes YOLOv13-ADR, an enhanced detection framework integrating Adaptive Deformable Convolution (ADConv) and a Nearest Neighbor Content Perception Recombination (NNCPR) module. ADConv applies a geometry-driven kernel permutation strategy to strengthen multi-scale feature representation, while NNCPR improves neighborhood-aware perception for modeling geometric deformations. A Focus-IoU loss function incorporating an anchor-quality perception mechanism is introduced to accelerate training convergence and improve bounding box regression precision. Additional optimizations include modifications to the DS-C3k2 module and upsampling strategy. Experiments on a wind turbine blade defect dataset demonstrate that YOLOv13-ADR achieves a 7.26-percentage-point improvement in mean average precision over YOLOv8n, with enhanced small-defect recognition and reduced localization errors, demonstrating improved detection precision and localization performance relevant to early fault detection and structural health monitoring of wind turbine blades. Full article

This study investigates the shear behavior and failure mechanisms of innovative hybrid structural beams fabricated by filling pultruded glass fiber-reinforced polymer (GFRP) box sections with waste rubber-reinforced concrete (RuC). Environmentally friendly concrete was produced by replacing natural aggregate with recycled tire-rubber fibers at proportions of 0%, 5%, 10%, and 15%. Twelve hybrid beam specimens were tested to evaluate the synergistic effects of rubber content and stirrup spacings of 16, 20, and 27 cm on shear capacity, ductility, and crack propagation. The experimental results revealed that the reference specimen (S16-0%) exhibited the maximum shear capacity of 154.41 kN and a brittle failure mode, while an increase in rubber content to 15%, combined with wider stirrup spacing, significantly reduced this capacity to a minimum of 96.89 kN (S27-15%). However, the 5% rubber replacement ratio achieved an optimal performance balance by preserving sufficient load-carrying capacity while enhancing flexural deformation and ductility, particularly in specimens with 16 cm stirrup spacing. Damage analysis demonstrated that longitudinal splitting cracks initiated in the mid-span tension zone at the bottom of the pultruded profiles, with final localized damage concentrated at the geometric corners of the box section. Crucially, the outer pultruded GFRP profiles provided substantial structural confinement, effectively mitigating the strength loss associated with high rubber incorporation and controlling the progression of sudden brittle failure. These findings highlight that combining pultruded GFRP profiles and optimized RuC offers a structurally viable and sustainable solution for modern infrastructure applications. Full article

Pump-as-Turbine (PaT) systems represent a cost-effective solution for hydraulic energy recovery in existing water networks and industrial processes. However, the prediction of their performance in turbine mode remains challenging, especially under off-design conditions where unsteady flow structures and internal losses strongly affect the machine efficiency. In this work, transient CFD simulations of a real industrial centrifugal pump operating as a turbine are performed using OpenFOAM and ANSYS CFX and compared with available experimental data. The investigated operating range extends from $0.7Q_{\mathrm{BEP}}$ to $1.3Q_{\mathrm{BEP}}$. A mesh independence analysis is first carried out using the Grid Convergence Index method, leading to the selection of a mid-size computational mesh as a compromise between accuracy and computational cost. The transient OpenFOAM results show close agreement with the ANSYS CFX predictions over the complete operating range. Both numerical frameworks reproduce the experimental hydraulic-efficiency trend and the location of the BEP, while systematic deviations in hydraulic head and mechanical power are mainly attributed to the geometrical and physical simplifications adopted in the common computational model. The local pressure coefficient monitored at the tongue region shows that both the mean pressure level and the fluctuation amplitude increase with flow rate, indicating stronger transient behavior under high-flow conditions. Beyond the global performance comparison, the flow field is analyzed using $Q_{crit}$ iso-surfaces, mean circumferential velocity, the swirl-intensity parameter $S_{\mathrm{int}}$, relative velocity fields at the PaT operational leading edge, and volute head-loss evaluation. The results show that part-load operation is characterized by strong outlet vortical structures and high residual swirl intensity, while the BEP region corresponds to reduced outlet rotational content. Under overload conditions, the outlet swirl remains limited, but the volute head loss increases significantly, becoming a dominant contributor to the efficiency drop. The study demonstrates that PaT performance cannot be interpreted from outlet swirl alone, but results from the combined effect of residual rotational structures, tongue-region unsteadiness, impeller incidence conditions, and volute dissipation. Full article

Hyperspectral image classification (HSIC) remains challenging because high-dimensional spectral signatures must be interpreted together with spatially coherent land-cover structures, particularly when labeled samples are limited. This paper presents a superpixel-based spectral–spatial HSIC method called SJSGR, which combines joint-sparse representation with graph Laplacian regularization. The HSI is first partitioned into homogeneous superpixel regions so that neighbouring pixels with similar spectral characteristics can be represented jointly rather than independently. For each superpixel, a similarity-aware weighting matrix is constructed between the training dictionary and the superpixel samples, encouraging the coefficient matrix to select more label-consistent and representative training atoms. To further preserve local manifold structure, graph Laplacian regularization is incorporated into the optimization objective, enforcing smooth and coherent representation coefficients among neighboring pixels within each superpixel. The resulting unified formulation integrates spectral correlation, spatial consistency, and local geometric structure, and is solved by the alternating-direction method of multipliers (ADMM). Classification is then performed by assigning each superpixel to the class with the minimum reconstruction error. Experiments are conducted on three real-world HSI datasets called Indian Pines, Pavia University and Fanglu to compare the proposed framework with several sparse representation and graph-based HSIC methods. Experimental results on these datasets reveal the capability of the proposed method, obtaining overall accuracies of 98.12%, 98.04%, and 98.26% under 10%, 1% and 1% labeled samples, respectively. Besides obtaining nearly 1% higher overall accuracy than the compared methods under these low-training-sample distributions, the SJSGR also provided better classification performance even under much more limited numbers of training samples. The findings suggest that superpixel-guided sparse representation with local manifold regularization is a promising direction for effective spectral–spatial HSIC. Full article

Masson pine (Pinus massoniana) is a core timber species in the subtropical regions of Southern China, and the precise monitoring of its growth status and phenotypic characteristics is crucial for forest resource management. To address the segmentation challenges caused by intertwined wood-leaf structures and severe occlusion, this study proposes PineSegNet, an end-to-end deep learning framework for fine-grained semantic segmentation of Masson pine point clouds. The framework adopts an encoder–decoder architecture. In the encoding stage, a Hierarchical Local–Global Aggregation (HLGA) module is introduced to capture multi-scale features through progressive downsampling. This design suppresses noise and enhances high-frequency geometric details of branches. In the decoding stage, a Boundary Refinement Unit (BRU) is designed to effectively curb feature diffusion during the interpolation process, significantly enhancing the clarity of category boundaries. Furthermore, this study develops a composite loss function with a dual-supervisory mechanism, namely the CE-Dice Composite Loss (CDC-Loss), to tackle semantic confusion caused by inter-class geometric similarity and the challenges of extreme sample distribution imbalance. Experimental results demonstrate that on both the self-collected dataset (GAOFENG) and the public dataset (FOR-instance), PineSegNet exhibits exceptional robustness and generalization capability, outperforming all evaluated semantic segmentation baselines under the unified experimental setting. This study provides a reliable technical solution for precision forest resource inventory and tree structural analysis within the framework of smart forestry. Full article

Accurate and robust vehicle localization is essential for autonomous driving. However, existing visual pose estimation methods often struggle in scenarios dominated by repetitive structures or sparse textures. These conditions lead to ambiguous predictions of 3D scene coordinates and a high proportion of structured outliers—erroneous predictions forming coherent clusters that deceive standard estimators. To address these limitations, this paper proposes DOU-Pose (Depthwise Over-parameterized U-shaped Pose estimation), a visual pose estimation framework built upon the Differentiable SAmple Consensus (DSAC)* pipeline. The core idea is to enhance the discriminative capability of scene coordinate regression through improved feature extraction. Specifically, we replace standard convolutional layers with Depthwise Over-parameterized Convolution (DO-Conv), which introduces auxiliary learnable depthwise kernels during training to enrich the representational capacity of the network, while allowing their fusion into a single kernel for inference. Furthermore, a U-shaped regression network with transposed convolutions is designed to preserve spatial details and strengthen fine-grained geometric reasoning. The entire pipeline is trained end-to-end by coupling dense scene coordinate prediction with a differentiable robust estimator. Extensive experiments demonstrate that DOU-Pose achieves competitive performance on public benchmarks and clear robustness improvements on the self-collected Campus-AV dataset, especially in repetitive and low-texture outdoor driving scenarios. Full article

Hydropower-generator rotors contain numerous closely spaced bolted joints, making full-coverage contact instrumentation impractical, while single-view vision methods are vulnerable to missing or ambiguous evidence during rotation. This study proposes a detector-guided multi-view visual monitoring framework that separates region-of-interest (ROI) localization from explicit geometric interpretation. On a simulated hydropower-generator rotor platform operating at 15 rpm, YOLO-family detectors localize candidate ${bolt}_{top}$, ${bolt}_{side}$, and ${star}_{marker}$ regions. Quality-retained top-view ROIs yield the image-space angular indicator ${\theta }_{img}$ from the relative orientation of nut-side and disk-side anti-loosening lines; side-view ROIs yield the pixel-space thread-exposure indicator $L_{px}$ from exposed-thread endpoints and, when marker geometry is sufficiently visible, an auxiliary angular cue. A star-shaped reference marker organizes accepted frame-level observations into approximate rotation intervals, while hierarchical checks of ROI completeness, endpoint availability, image quality, geometric plausibility, and temporal membership retain both usable evidence and explicit rejection reasons. YOLO11n achieved precision 0.9987, recall 1.0000, mAP50 0.9950, and mAP50–95 0.7798 for laboratory ROI localization. After geometric screening, evidence availability was 16.7% for the top-view branch and 76.3% for the side-view branch. In a supplementary 169-image operational field subset, the principal ROI model achieved precision 0.9782, recall 0.9942, mAP50 0.9946, and mAP50–95 0.7948. The field results support appearance-level localization under complex rotor-bolt conditions, and the framework provides a traceable, reliability-aware basis for organizing, screening, and interpreting multi-view evidence in hydropower-generator rotors and similar rotating structures. Full article

Portable handheld three-dimensional (3D) scanners have improved facial surface measurement workflows in plastic and reconstructive surgery and have potential applications in preoperative planning, postoperative assessment, and customized implant design. However, the relative agreement of low-cost passive stereophotogrammetry for nasal surface reconstruction remains insufficiently characterized when compared with handheld structured-light scanning. This study evaluated the feasibility and relative surface-deviation performance of two passive stereophotogrammetry-based systems for nasal 3D surface reconstruction in healthy participants. Twenty-five healthy participants were enrolled, and nasal 3D models were reconstructed using Artec Eva, COLMAP, and Agisoft Metashape. COLMAP- and Agisoft Metashape-derived models were aligned to the corresponding Artec Eva models in MeshLab. Geometric deviations were quantified using root mean square (RMS) surface deviation, mean surface distance (MSD), and Hausdorff distance (HD), and regional discrepancies were visualized using surface-distance heat maps. Both passive pipelines produced deviations generally within previously reported ranges for facial or nasal 3D imaging, with median RMS and MSD values below 0.5 mm. Hausdorff distance showed larger local deviations, particularly in the alar and perinasal regions, and was significantly lower for Agisoft Metashape than for COLMAP. These findings suggest that low-cost passive stereophotogrammetry may support standardized nasal surface measurement under controlled acquisition conditions. Full article

Three-dimensional (3D) point cloud change detection is essential for urban monitoring and environmental analysis, yet existing methods mainly rely on point-wise semantic differences and overlook change–unchange boundaries and object edges—key geometric cues for precise localization, especially for subtle or gradual changes. To address this, we propose GSANet, a Geometric Structure-Aware Siamese Network that explicitly integrates boundary and edge priors into sampling and feature learning. First, a Boundary-Aware Subsampling (BAS) strategy preserves key points near change boundaries while reducing redundancy, and a Boundary-Aware Binary Cross-Entropy (BA-BCE) loss assigns higher supervision weights to boundary points, enhancing learning in ambiguous regions. Second, an Edge-Aware Siamese Network captures robust local shapes by embedding edge priors into feature extraction, incorporating Edge-Aware Adaptive Graph Convolution, Edge-Aware Downsampling, and Cross-Attention Upsampling to maintain structural consistency across temporal branches. Additionally, a Difference Enhancement Module (DEM) amplifies feature discrepancies between bitemporal point clouds, improving sensitivity to subtle changes. Extensive experiments on a street-level dataset and urban dataset show our method outperforms state-of-the-art approaches. Full article

Corrosion remains one of the principal causes of degradation in metallic structures across a wide range of industrial sectors. Although visual inspection is routinely employed for preliminary corrosion assessment, its effectiveness depends heavily on operator experience and subjective interpretation. This work introduces CorrQuant, a web-based computer vision platform designed to transform qualitative corrosion images into quantitative measurements of corrosion extent and morphology. The proposed methodology processes images acquired with conventional mobile devices and integrates geometric calibration using a reference coin, perspective correction, adaptive image enhancement through Contrast Limited Adaptive Histogram Equalization (CLAHE), multi-descriptor feature extraction, and consensus-based corrosion segmentation. The detected corrosion regions are subsequently quantified to determine corrosion area, surface coverage, spatial distribution, morphological descriptors, and corrosion intensity maps. The methodology was verified using an aluminum specimen with a known corrosion area of 143 mm² under both controlled illumination and optical stress-test conditions. Under standard acquisition conditions, corrosion-area estimation accuracies exceeding 90% were achieved. Additional evaluations under red illumination, fisheye, blur, and kaleidoscope distortions demonstrated that the proposed framework is considerably more sensitive to degradation of local image information than to variations in illumination spectrum. These results demonstrate the robustness of the proposed multi-descriptor voting strategy while defining the operational limits of the platform under challenging image acquisition conditions. Full article

Risk assessment of Unmanned Aerial Vehicle (UAV) path planning is a crucial step towards ensuring a safe UAV operation. However, achieving a reasonable risk assessment of UAV flight paths remains challenging because it involves multiple responsibilities and processes that extend beyond a single agency or organization. Additionally, it must balance various complex factors and data from social, technical, political, and economic sources. Most existing works on flight path planning evaluate flight risks at a global level and generalized geometric representations of the UAV operating environment with respect to the current applicable UAV regulations. However, geometric information alone does not provide sufficient insight for comprehensive and safe path planning. Therefore, there are other ideas and concepts for using semantic data to characterize objects, obstacles and actions in the UAV environment. Incorporating semantic information into path planning enables more meaningful scene descriptions and a better representation of relevant constraints, obstacles within UAV environment that may influence the safety level of UAV operation, and the relevant risk assessment process. In this paper, we propose the development of methods and frameworks integrated into a flight planning prototype designed to generate safe two-dimensional UAV routes within a local, fine-grained planning context. The prototype incorporates safety considerations to enable a comprehensive assessment of UAV operational risks. It leverages both geometric and semantic datasets to characterize objects and obstacles within the UAV environment. These datasets are processed and stored in a relational database to support structured access and long-term usability. All concepts and experiments were implemented using datasets from a study area in the city of Brunswick, Germany. Full article

This study proposes a novel friction-based beam-to-column shear connection designed to behave as a nominally pinned joint while avoiding any perforation of the connected members. The connection relies on frictional resistance to transfer shear forces, enabling full reversibility and preserving the integrity of the structural elements for future reuse. A comprehensive design methodology is first introduced, addressing key parameters such as clamping force, friction coefficient, and slip resistance. Subsequently, an extensive numerical investigation is carried out using refined finite-element models, i.e., considering multiple geometric configurations and loading conditions. The local behaviour of the connection is hence assessed in terms of stiffness, strength, and slip capacity. Results show that—with proper sizing—plastic deformation localises in the beam while the joint remains elastic and slip is limited, confirming the conservativeness of the approach. The joints behave as nominally pinned in terms of resistance while showing moderate stiffness. Derived findings highlight the feasibility of adopting friction-based, non-invasive connections as a viable alternative for circular steel construction, contributing to the ongoing transition toward more sustainable structural systems. Full article

The quantum–classical transition is one of the most frequently invoked concepts in modern physics. Yet the notion of classicality itself is far from unique. Depending on the physical context, classical behavior may be associated with decoherence, the semiclassical limit, thermodynamic averaging, suppression of correlations, emergence of collective order, or geometric simplification of statistical state space. These viewpoints are often presented as if they described a single phenomenon, although they emphasize different physical mechanisms and different operational criteria. In this article, we examine the principal notions of classicality that appear across quantum theory, statistical physics, condensed matter physics, and information geometry. We compare the corresponding mechanisms of classical emergence and analyze the physical quantities commonly used to characterize them, including coherence, entanglement, fluctuations, correlation length, Fisher information, statistical complexity, and information-geometric curvature. We argue that many apparently distinct routes toward classical behavior share a common structural feature: a reduction of effective fluctuation freedom (REFF). From this perspective, classicality may be interpreted as an emergent regime in which the accessible fluctuation manifold becomes progressively constrained, stabilized, or geometrically simplified. This viewpoint naturally unifies decoherence, semiclassical localization, thermodynamic averaging, decorrelation, and collective organization within a common conceptual framework. Rather than representing a unique physical process, the quantum–classical transition appears as a family of related mechanisms through which complex quantum fluctuation structure gives rise to effective macroscopic classical behavior. Full article

Monocular depth estimation aims to recover dense 3D scene geometry from a single RGB image and plays an important role in autonomous driving, robotic perception, augmented reality, and 3D reconstruction. Although Transformer-based dense prediction models have achieved strong performance, existing DPT-Hybrid frameworks still suffer from three limitations: insufficient local geometric modeling in shallow stages, inadequate cross-scale fusion for preserving fine structures, and training objectives that only weakly constrain structural consistency. To address these issues, we propose a structure-aware enhanced DPT-Hybrid framework. First, a geometry-enhanced encoder introduces lightweight depth-wise separable convolution branches into shallow Transformer stages to better capture local edge and texture cues while preserving global contextual modeling. Second, a Structure-Aware Cross-Scale Gated Attention Fusion (S-GAF) module is proposed to improve decoder-side feature aggregation by jointly modeling channel-wise and spatial importance with an auxiliary RGB-gradient input. Third, joint structure–geometric consistency loss combines scale-invariant logarithmic loss, gradient consistency loss, and edge-focused loss to improve pixel-level accuracy, geometric plausibility, and boundary sharpness. Experiments on NYUv2 and KITTI demonstrate that the proposed method achieves lower single-run error metrics than the controlled DPT-Hybrid baseline under the evaluated settings. On NYUv2, our method achieves an absolute relative error (AbsRel) of 0.099 and an RMSE of 0.334. On KITTI, it achieves an AbsRel of 0.058 and an RMSE of 2.455. The proposed method introduces only modest additional complexity while producing more accurate and structurally sharper depth predictions. Full article

Early-stage alfalfa navigation faces challenges like low plants, narrow rows, and weed interference, causing camera–LiDAR colored point clouds to suffer from sparsity, discontinuous boundaries, and varying illumination. Standard point-level semantic segmentation struggles to support stable crop row allocation and navigation line fitting under these conditions. To address this, we propose RACG-RandLA, a multi-output row-aware color-geometric point cloud segmentation model. Pseudo-labels (crop, background, ‘ignore’) are generated using color and spatial priors, alongside transverse offset and point-level confidence labels for crop points. Built on RandLA-Net, the multi-task network simultaneously outputs crop semantics, transverse offsets, and confidences. It features a color-geometry residual fusion module that adaptively integrates 3D geometric and RGB/ExG features via zero-initialized scaling to handle complex lighting and missing data. Additionally, a late Row-cued Local Feature Aggregation (LFA) module embeds longitudinal continuity and transverse offset constraints into deep layers, effectively mitigating cross-row feature aliasing. During inference, a multi-output pipeline utilizes these predictions for reliable crop point filtering, center refinement, and navigation line fitting. Experiments show that the xyzrgb_exg input achieves an optimal balance between accuracy and conciseness. RACG-RandLA achieves a Test IoU of 0.8590, outperforming PointNet variants, and secures the highest Row Count Accuracy of 0.8761. Furthermore, it reduces lateral navigation jitter to 0.0241 m while maintaining an 85.04% success rate. Ultimately, the proposed method demonstrates a superior balance of semantic accuracy, structural consistency, and navigation stability, providing a robust perception solution for agricultural robots. Full article