Search Results (133)

Accurate finite element models are essential for structural condition assessment, yet nominal material properties and idealized boundary conditions can produce systematic discrepancies between numerical and measured dynamics. This study proposes a physics-constrained sparse response-surface framework that combines Elastic Net basis selection, mechanically prescribed monotonicity, adaptive sample enrichment, and identifiability-aware uncertainty assessment within a transparent finite element model-updating procedure. A scaled steel truss was tested using millimeter-wave radar, and the first three vertical natural frequencies were identified by stochastic subspace identification. The resulting sparse polynomial surrogate was independently validated before bounded inversion and ANSYS back-substitution. The mean frequency error decreased from 5.55% to 0.82%. Jacobian and bootstrap analyses further showed that several combinations of material and boundary parameters can reproduce similar modal responses, so the updated parameters are best interpreted as a coupled equivalent calibration state rather than unique direct measurements. The proposed framework therefore improves physical consistency and computational efficiency while explicitly retaining the uncertainty associated with weakly identifiable parameter directions. Full article

Long-span cable-stayed bridges undergo substantial changes in mass distribution, boundary conditions, cable forces, and load paths during erection, yet their stage-dependent free-vibration characteristics are less documented than those of completed bridges. This study establishes a refined finite element (FE) model of the Liulu Yongjiang Extra-large Bridge and evaluates the frequency evolution over 26 construction stages. Detailed modal interpretations are provided for the maximum double-cantilever, maximum single-cantilever, and completed-bridge configurations. The subspace iteration eigensolver is used for modal extraction, and ambient-vibration measurements at the three representative stages provide an independent frequency check. The calculated fundamental frequencies are 0.399, 0.417, and 0.436 Hz for the three configurations, respectively. The governing mode changes from antisymmetric vertical girder bending at the maximum double-cantilever stage to lateral girder bending at the maximum single-cantilever and completed-bridge stages. Across CS1-CS26, cantilever extension generally reduces the governing frequency, cable installation produces a diminishing vertical-stiffening effect, and the modeled 390 t form-traveler mass lowers both lateral and vertical frequencies. Field tests indicate that the measured frequencies at three stages deviate less than 9.6% from the FE values, confirming the reliability of the model. The revealed evolution laws and influencing mechanisms of dynamic characteristics during construction provide theoretical support for vibration control and safety assurance of similar bridges. Full article

Colorectal polyp segmentation requires accurate boundary delineation, but adapting large vision foundation models to endoscopic images can be computationally expensive. This study investigates whether the Segment Anything Model (SAM) can be specialized for polyp segmentation by updating only a small attention subspace. We propose PolypSAM-Lite, which freezes the SAM vision backbone and applies low-rank reparameterization only to the fused Query–Key–Value attention projections. The model uses bounding-box prompts, binary cross-entropy plus Dice loss, AdamW optimization, and evaluation on Kvasir-SEG with external testing on CVC-ClinicDB and ETIS-LaribPolypDB. PolypSAM-Lite updates 4.2 million parameters and achieves a Dice Similarity Coefficient of 0.9507 on Kvasir-SEG, compared with 0.8804 for zero-shot SAM. External Dice scores are 0.9271 on CVC-ClinicDB and 0.9198 on ETIS-LaribPolypDB, indicating cross-dataset generalization under domain shift. These results suggest that QKV-restricted low-rank attention reparameterization can provide an efficient and effective strategy for adapting SAM to colorectal polyp segmentation without full-backbone fine-tuning. Full article

Accurate channel estimation in frequency-selective multiple-input multiple-output orthogonal frequency division multiplexing (MIMO-OFDM) systems requires balancing pilot overhead, reconstruction accuracy, and computational cost. This paper presents a reproducible compressed sensing benchmark for sparse delay-domain channel estimation with reduced pilot observations. Its novelty is not the invention of Orthogonal Matching Pursuit (OMP), Compressive Sampling Matching Pursuit (CoSaMP), or Subspace Pursuit (SP), but the construction of a transparent and auditable evaluation protocol in which all estimators operate on the same channel realizations, sensing matrices, pilot budgets, signal-to-noise ratios (SNRs), stopping rules, and Monte Carlo trials. The framework explicitly defines the underdetermined observation model, the per-link 4 × 4 MIMO interpretation, the minimum-norm least-squares (LS) baseline, the identity-prior linear minimum mean-square error (LMMSE) baseline, the oracle-known sparsity assumption, uncertainty reporting, runtime protocol, and the mapping from delay-domain estimates to link- and subcarrier-domain quantities. OMP, CoSaMP, SP, LS, and LMMSE are evaluated for a 128-element delay dictionary, five active taps, sampling ratios from 0.10 to 0.70, SNRs from 0 to 30 dB, and 80 independent trials per operating point. The results show that sparse recovery exploits the assumed delay-domain sparsity more effectively than non-sparse baselines in the underdetermined regime, while pilot density remains a dominant factor in support identification and reconstruction error. The accompanying Python human–machine interface (HMI) produces confidence-aware metrics and publication-ready figures, enabling exact repetition of the benchmark and controlled extension to more realistic channel models. The conclusions are limited to simulation-based algorithmic evidence and define a direct pathway toward standardized-channel, software-defined radio (SDR), and measured radio-frequency (RF) validation. Full article

Hyperspectral image denoising must reconcile global spectral coherence with spatially heterogeneous scene content. This paper presents Adaptive Heterogeneity-Aware Tensor Decomposition (AHTD), a refinement module that augments a global Tucker initialization with heterogeneity-guided local tensor shrinkage. The method estimates a global spectral subspace, detects heterogeneous regions via combined variance and edge analysis, and applies adaptive weighted singular-value shrinkage modulated by regional patch complexity. To isolate the effect of the heterogeneity-aware selection mechanism itself, experiments employ a controlled internal ablation protocol comparing three pipeline variants under identical noise realizations, ranks, and parameter settings on the Pavia_80, Indian Pines corrected, and Salinas corrected benchmarks under synthetic mixed noise ($\sigma =0.03$ Gaussian with stripe and impulse noise): a global Tucker baseline, a dense full-local refinement variant, and the proposed selective AHTD. AHTD achieves peak signal-to-noise ratio (PSNR) gains of 0.35–0.40 dB over the global Tucker baseline while maintaining or improving structural similarity index (SSIM), spectral angle mapper (SAM), and relative dimensionless global error in synthesis (ERGAS), at approximately threefold lower runtime than dense full-local refinement, demonstrating its value as a computationally efficient, interpretable refinement stage for tensor-based hyperspectral processing. We note that matched comparisons against externally published denoising methods are not included in this study; these results establish the benefit of heterogeneity-aware selective refinement within the proposed Tucker-based pipeline. Full article

Finger vein recognition (FVR) has significant potential in biometrics due to its high accuracy and intrinsic liveness detection capabilities. However, the increasingly stringent privacy regulations have presented severe data security challenges for traditional centralized training. While federated learning (FL) mitigates these privacy concerns through a decentralized training paradigm, conventional FL algorithms that seek a single global model experience significant performance degradation on non-independent and identically distributed (Non-IID) data in real-world cross-institutional deployments. This degradation stems primarily from a dual-heterogeneity issue that involves domain shift caused by hardware discrepancies across acquisition devices, and label skew resulting from nonoverlapping user identities. To address this dual-heterogeneity challenge, we propose a personalized federated learning framework driven by hierarchical parameter decoupling and subspace metric. First, we designed a hierarchical parameter decoupling architecture. Macroscopically, the architecture retains the classifier locally to isolate label heterogeneity; microscopically, it introduces an additive parameter decomposition that decouples the feature extractor on a global full-rank basis (to capture domain-invariant semantics, namely, the shared physiological vein topologies) and a local low-rank adapter (that accommodates device-specific characteristics, such as hardware-induced noise and illumination discrepancies). Furthermore, we propose a subspace similarity matching strategy based on principal angles on the Grassmann manifold. By exploiting the geometric properties of low-rank projection matrices, this strategy accurately quantifies the underlying distribution discrepancies among clients to guide personalized weighted aggregation. Extensive experiments on six public finger vein datasets demonstrate that the proposed framework significantly improves the overall recognition performance and mitigates performance degradation caused by data heterogeneity. Full article

Wild carrot (Daucus carota subsp. carota), the wild relative of cultivated carrot, is globally identified as an invasive weed that threatens hybrid carrot seed production through natural cross-pollination, resulting in compromised genetic purity. Manual identification across the large areas required to ensure genetic purity in carrot seed crops is impractical. Remote sensing offers an alternative; however, morphological similarities among wild carrot, cultivated carrot, and common weeds hinder reliable detection. Early identification, however, remains essential for preventing genetic contamination. This study evaluated leaf-level hyperspectral reflectance spectroscopy (400–2450 nm) with machine learning to discriminate wild carrot from hybrid carrots, parental lines, and 19 co-occurring weed species. Spectral data from 266 wild carrot plants across three New Zealand sites and six weeks (5–10 weeks after emergence) showed negligible spatial effects (R² = 0.034–0.055, pseudo-F = 1.46–2.39, p > 0.05) and moderate temporal variation (R² = 0.136–0.151, pseudo-F = 5.48–6.17, p < 0.001), indicating broadly stable spectral signatures suitable for model generalization. Savitzky–Golay filtering, with min–max normalization outperformed SNV, yielding high full-spectrum accuracies for wild carrot vs. other species (90.35%, κ = 0.80), wild carrot vs. weeds (96.03%, κ = 0.92), and a multi-class model (90.79%, κ = 0.88). After removing atmospheric water-absorption bands to follow airborne sensing, reduced-band models based on airborne-compatible wavelengths maintained strong performance, including 89.40% accuracy (κ = 0.79) for wild carrot vs. weeds using a 20-band Subspace Discriminant model (400–402, 527, 705–720 nm). These findings demonstrate that stable wild carrot spectra and carefully selected visible and red-edge bands can underpin cost-effective UAV/UGV-mounted hyperspectral or multispectral sensors for site-specific wild carrot management. Full article

Controllable symbolic music generation must preserve a reference melody while remaining responsive to style prompts. Existing hierarchical diffusion systems typically reuse a shared condition vector across harmony, rhythm, and timbre stages, which can entangle stylistic factors and weaken melody preservation. We present HCDMG++, a hierarchical diffusion framework that addresses these two limitations through stage-aware style routing and differentiable melody regularization. The routing module uses a residual multi-layer perceptron (MLP) with zero-initialized scalar gates to project text-derived style embeddings into harmony-, rhythm-, and timbre-specific subspaces, whereas the regularization branch aligns soft pitch histograms and contour trajectories with the conditioning melody during training without breaking the differentiable computation graph. We evaluate the integrated system on a 384-sample benchmark covering four melodies, eight styles, four random seeds, and three denoising budgets, supplemented by a matched legacy-compatible reference and inference-time component ablation that contrasts legacy behavior, silenced gates, an automated uniform gamma routing sweep, and the full forward pass. HCDMG++ produces valid four-track outputs in all 384 runs, reaches a peak pitch histogram similarity score of 0.508 under a 64-step budget, and improves pitch histogram alignment over Legacy-HCDMG by roughly two orders of magnitude on the matched slice, while attaining a positive Fisher-style style separability score where the legacy benchmark is too sparse to support one. These results indicate that stage-specific conditioning and differentiable structural guidance jointly improve controllability in symbolic music diffusion, while also exposing the remaining limitations in long-form generalization and perceptual validation, which motivate the future work outlined at the end of this paper. Full article

Decomposition of nonlinear, nonstationary multicomponent signals remains challenging for existing decomposition strategies, including frequency-based, data-driven, and subspace methods, which can suffer from mode mixing, leakage across components, and unreliable isolation of transients. Motivated by this gap, this study proposes Orthogonal Self-Similarity Decomposition (OSSD), which exploits a self-similarity structure in delay-embedded orbit geometry so that temporal organization, rather than spectrum alone, guides component construction. OSSD-Basic introduces three algorithmic novelties within a single pipeline: (1) an adaptive proxy-correlation band merging on the delay axis, (2) a dominant-component cascade that prevents energy-dominant carriers from masking weaker components, and (3) a double MGS + LS reprojection that collapses the inter-mode orthogonality index to numerical zero, regardless of merging and pruning operations. Synthetic experiments with known ground truth show that OSSD-Basic provides a parsimonious four-mode representation with exact inter-mode orthogonality (OI = 9.4 × 10⁻¹⁸), the highest reconstruction SNR among the evaluated baselines (27.14 dB), and the highest ground-truth diagonal correlation sum (3.038) among the tested methods, while using two fewer modes than EMD, VMD, and SSA. Daily streamflow forecasting on a U.S. Geological Survey discharge record further shows that augmenting OSSD-derived inputs with fractal descriptors and fractional-order differencing features yields progressive accuracy gains over the AR-ANN baseline, with R² improving from 0.855 to 0.915 at one-step-ahead and from 0.388 to 0.699 at four-step-ahead forecasting in the single-input setting, within a single-station case study on USGS 01554000. Overall, OSSD-Basic offers an interpretable multiscale decomposition with guaranteed inter-mode orthogonality and a structured feature pathway for oscillatory–transient mixtures. Full article

In the classical Stochastic Subspace Identification (SSI) method, mode selection is primarily based on frequency stability, damping stability, and mode shape similarity using the Modal Assurance Criterion (MAC). However, these criteria are often insufficient for reliable modal identification in high-noise environments. This study advances beyond the classical approach by introducing a multi-criteria optimization framework for mode evaluation. In addition to the conventional frequency and damping assessments utilized in the classical SSI method, the proposed approach incorporates a range of supplementary structural metrics. These include Density, Cosine Similarity Difference (CSD), Damping Stability (DS), Spatial Roughness (SR), Mode Shape Complexity (MSC), Signal Energy Coherence (SEC), and Normalized Modal Difference (NMD). These metrics are computed within specifically optimized windows on the stabilization diagram. By integrating spatial, phase, and energy-based characteristics of mode shapes alongside traditional metrics such as the MAC, the method enables a more comprehensive and robust mode selection process that surpasses the limitations of relying solely on frequency and damping stability. Compared to the classical SSI, the optimized window approach provides a significant advantage by enabling the reliable selection of consistent modes by considering the continuity and multi-criteria coherence of modes across window transitions. As a result, the elimination of noise modes and the reliable separation of structural modes are established on a more systematic basis. To achieve this, a two-stage optimization strategy is implemented: the first stage determines the optimal frequency window width and minimum mode count threshold, while the second stage utilizes a Multi-Criteria Decision Making (MCDM) framework based on the Technique for Order Preference by Similarity to Ideal Solution (TOPSIS) algorithm to assign optimized weights to the structural metrics and rank the candidate windows accordingly. As a result, the ideal frequency window is identified based on its TOPSIS score and subsequently validated using the MAC, confirming that the selected window corresponds to reliable structural modes. The framework is validated using long-term in situ measurements from a Roller Compacted Concrete (RCC) dam operating under significant environmental and operational noise. The dataset comprises continuous, high-resolution (200 Hz) vibration recordings collected between 1 July 2023 and 30 October 2024. While the calendar duration is limited to several weeks, the uninterrupted 24 h measurements yield a high-density time-series dataset with substantial information content, enabling a statistically meaningful and robust evaluation of modal identification performance under real-world and noisy conditions. The results reveal that relying solely on traditional selection criteria such as pole density and the MAC can often lead to the identification of spurious modes, particularly in noisy environments. In contrast, the proposed TOPSIS-based multi-criteria decision-making framework incorporates a broader range of structural indicators, balancing frequency, damping, spatial, and energy-related metrics to enhance the consistency and reliability of mode selection. This approach proved effective even under high-noise conditions, successfully distinguishing true structural modes from artificial ones. Application of the TOPSIS method to RCC dam data revealed consistent fundamental frequencies at approximately 5–10 Hz, 10 Hz, and 15 Hz, confirming its robustness and suitability for complex structural monitoring tasks. Full article

Existing multi-view clustering approaches based on matrix factorization often fail to jointly capture global high-order correlations and local view-specific characteristics, and they typically suffer from instability in generating final clustering labels. To overcome these limitations, this paper presents a multi-view subspace clustering method termed dual-tensor constrained multi-view subspace clustering (DTCMVSC). Specifically, for each view, we learn an independent latent representation matrix, a projection matrix, and a basis matrix. The latent representations and projection matrices are stacked into third-order tensors, upon which tensor nuclear norm regularization is imposed to simultaneously exploit consensus structures and complementary information across views. Additionally, a consensus regularization term and adaptive view weights are introduced to align the latent representations of different views toward a unified consensus subspace. The resulting optimization problem is efficiently solved under the ADMM framework, after which a similarity matrix is constructed from the consensus representation and spectral clustering is performed to obtain the final labels. Experimental evaluations on six benchmark datasets demonstrate the superiority of DTCMVSC. Specifically, it achieves an ACC of 86.10% on CMU and an NMI of 94.17% on ORL, surpassing even the lowest-performing state-of-the-art baselines by 63.08 and 18.53 percentage points, respectively. Full article

Few-shot fine-grained image classification aims to recognize subcategories with high visual similarity using only a limited number of annotated samples. Existing metric learning-based methods typically rely solely on spatial-domain features. Confined to this single perspective, models inevitably suffer from inherent texture biases, entangling essential structural details with high-frequency background noise. Furthermore, lacking cross-view geometric constraints, single-view metrics tend to overfit this noise, resulting in structural instability under few-shot conditions. To address these issues, this paper proposes the Dual-Subspace Network (DSNet). Specifically, DSNet utilizes the discrete cosine transform (DCT) and a low-pass filtering mechanism to explicitly isolate low-frequency global structural components from spatial features, thereby suppressing background interference. Truncated Singular Value Decomposition (SVD) is employed to construct independent, low-rank linear subspaces for both spatial texture and frequency structural features. An adaptive gating mechanism is designed to dynamically fuse the projection distances from these dual views. This strategy leverages the structural stability of the frequency subspace to prevent the spatial subspace from overfitting to background features. Extensive experiments on four benchmark datasets—CUB-200-2011, Stanford Cars, Stanford Dogs, and FGVC-Aircraft—demonstrate that DSNet exhibits excellent classification performance and robustness, achieving highly competitive results compared to existing metric learning algorithms. Complexity analysis further confirms that the proposed network achieves a favorable balance between high accuracy and computational efficiency, providing an effective new paradigm for few-shot fine-grained visual recognition. Full article

As an innovative metaheuristic algorithm, Genghis Khan Shark Optimization (GKSO) faces challenges, including a tendency towards local optima and poor convergence speed and accuracy. To mitigate these limitations, an improved Genghis Khan shark optimizer (IGKSO) is proposed in this paper. A population partitioning method based on cosine similarity and fitness is introduced, where individuals are strategically assigned to different evolutionary phases: Disadvantaged populations are responsible for the foraging stage. By contrast, advantaged populations dominate the moving stage. In the moving stage, the base vector is randomly selected from multiple candidates, which ensures the evolutionary direction of the population while maintaining its diversity. An adaptive step-size mechanism is introduced to avoid boundary overflow problems. A subspace method is employed to prevent diversity loss during foraging. Additionally, in the hunting stage, a novel opposition-based learning strategy is proposed to moderate the tendency of converging to suboptimal solutions. Furthermore, during the self-protection phase, a criterion for assessing the diversity of the whole population is employed to monitor and supplement diversity in real time. The results of the CEC2017 and CEC2019 benchmark test sets reveal that IGKSO exhibits substantial advantages over the GKSO algorithm and eight other high-performance algorithms in terms of convergence speed and accuracy. Full article

Longan is an important economic fruit in tropical and subtropical regions, whose harvesting primarily relies on manual labor. Automated longan harvesting is key to improving the industry’s economic benefits but faces core challenges: mature pericarp is highly similar in color to fruiting mother branches, plus dense branches and severe leaf occlusion, leading to difficult cluster detection and fruiting branch segmentation. Herein, we propose a picking point localization method named PLYS-Longan integrating three customized core modules: Dynamic Convolution, Convolutional Gated Linear Unit (CGLU), and Dynamic Hyperbolic Tangent Activation (DYT) are introduced into YOLongan module to enhance the model’s ability to detect longan clusters. For SELongan module, Depthwise Over-parameterized Convolution (DO-Conv) and Ultra-light Subspace Attention (ULSA) are adopted to improve main branch segmentation precision. The PCLongan module then performs morphological erosion on the segmentation masks and calculates centroids to precisely determine the picking points. Experimental results show that the improved model achieves a mAP@50 of 90.1% (3.3% higher than baseline model) in object detection and a mIoU of 77.24% (1.75% improvement) in semantic segmentation, outperforming the various model significantly. This study provides an efficient and robust solution for longan picking point localization, laying a solid foundation for subsequent automated harvesting. Full article

Support Vector Machine (SVM) is a popular kernel-based method for data classification that has demonstrated high efficiency across a wide range of practical applications. However, SVM suffers from several limitations, including the potential failure of the optimization process, especially in high-dimensional spaces; the inherently high computational cost; the lack of a systematic approach to multi-class classification; difficulties in handling imbalanced classes; and the prohibitive cost of real-time or dynamic classification. This paper proposes an alternative method, referred to as Kernel-based Optimal Subspaces (KOS), which belongs to the family of kernel subspace methods. Mathematically similar to Kernel PCA (KPCA), KOS achieves performance comparable to SVM while addressing the aforementioned weaknesses. The method is based on computing the minimum distance to optimal feature subspaces of the mapped data. Because no optimization process is required, KOS is robust, fast, and easy to implement. The optimal subspaces are constructed independently, enabling high parallelizability and making the approach well-suited for dynamic classification and real-time applications. Furthermore, the issue of imbalanced classes is naturally handled by subdividing large classes into smaller sub-classes, thereby creating appropriately sized sub-subspaces within the feature space. Full article