Search Results (1,196)

Text-based person search (TPS) suffers from cross-modal informational skewness: pedestrian images are high-dimensional and redundancy-prone, while textual descriptions are sparse, incomplete, and sometimes inaccurate. To address the low alignment accuracy and poor robustness caused by the inherent uneven information distribution of visual and textual modalities in TPS, this paper proposes a unified Partition-based Information Rebalancing (PaIR) framework to realize balanced optimization and precise alignment of cross-modal information from both global content and local part dimensions. The framework adopts the CLIP dual-modal encoder for basic feature extraction and constructs a parallel global–local dual representation system to compensate for the lack of fine-grained spatial information in single global features. To eliminate modal redundancy and noise interference, a dual-modal noise suppression module is designed to filter invalid redundant information through visual foreground–background separation and textual token weight screening, while introducing adversarial constraints and orthogonal constraints to purify effective features. On this basis, a part balance alignment module is built to complete human semantic part decomposition and soft matching alignment for dual-modal features. Aiming at the common part semantic missing problem in textual descriptions, a visual part correlation affinity matrix is utilized for semantic associative completion to balance the information density of dual modalities. Finally, a global–local joint alignment strategy integrates hierarchical features and bidirectional cross-modal attention interaction to eliminate global–local semantic discontinuity and enhance fine-grained cross-modal matching capability. Extensive experiments on three public benchmarks demonstrate that PaIR consistently improves multiple baselines. Full article

Accurate and rapid detection of citrus leaf diseases is important for early diagnosis, precision orchard management, and the reduction of economic losses in citrus production. Automatic detection remains difficult because early lesions are often small and irregular. Several disease categories also share similar visual appearances, and localization is easily affected by veins, shadows, and cluttered backgrounds. To address these task-specific challenges, we propose HGSM-YOLO, where HGSM denotes the coordinated use of heterogeneous convolution, a GSConv-based slim neck, and multi-scale dilated local attention. The framework is built on YOLO11n because its 2.59 M-parameter and 6.4 GFLOP design provides a stringent compact baseline for edge-oriented improvement. The method follows a hierarchical design: C3k2-HetConv preserves lesion edges and local morphology in the backbone; the GSConv-based slim neck reduces part of the feature fusion cost; and an MSDA module in the high-resolution P3 branch enhances the context of small lesions. Following model selection on the validation split, the final locked models were evaluated once on the held-out test split, with HGSM-YOLO reaching 77.5% precision, 66.8% recall, 71.7% F1-score, 70.8% mAP@0.5, and 44.2% mAP@0.5:0.95, compared with 68.5%, 61.5%, 64.8%, 66.0%, and 40.2% for YOLO11n. A stratified outer five-fold cross-validation further yields 71.0% ± 1.4% mAP@0.5 and 44.4% ± 1.1% mAP@0.5:0.95 for HGSM-YOLO, versus 65.9% ± 1.1% and 40.2% ± 0.9% for YOLO11n. On the independent 1871-image citrus-leaf-disease-2 dataset, retraining under the same protocol gives 94.4% mAP@0.5 for HGSM-YOLO versus 92.2% for YOLO11n and 93.1% for the public Roboflow YOLOv11 reference model. The complete HGSM-YOLO architecture uses 7.2 GFLOPs, 2.82 M parameters, and runs at 90.9 FPS on the RTX 4090, compared with 6.4 GFLOPs, 2.59 M parameters, and 110.1 FPS for the baseline. Thus, the contribution provides a recall- and localization-oriented accuracy–efficiency trade-off rather than universal superiority in every individual metric. Full article

Small object detection in UAV remote sensing imagery plays a crucial role in applications such as infrastructure inspection, disaster assessment, and precision agriculture, where targets of interest frequently occupy fewer than $32\times 32$ pixels under large ground sampling distance variation and complex cluttered backgrounds. Existing methods still face three main challenges in UAV small-object detection: fine-grained detail loss caused by repeated downsampling, feature inconsistency during cross-scale fusion, and unstable boundary regression in densely distributed aerial scenes. To address these issues, this paper proposes HSAR-DETR, a detection framework that jointly improves hierarchical feature representation, cross-scale refinement, and geometry-aware localization. Specifically, a Hierarchical Enhancement Network (HENet) is introduced to preserve shallow spatial details while strengthening deep semantic-context representation. A Dual-Stream Feature Refinement module (DSFR) is designed at the P4-to-P3 fusion stage, combining spatial-domain structural modeling with frequency-domain phase refinement to improve cross-scale feature consistency. A Coordinate-Guided Adaptive Convolution module (CGAC) is further deployed before the detection head, converting coordinate-guided offset magnitudes into modulation weights for adaptive feature recalibration and improved localization stability. In addition, a conventional high-resolution P2 detection branch is incorporated to enhance small-object representation. Experimental results on the VisDrone, RSOD, and TinyPerson datasets demonstrate improved detection performance. On the VisDrone validation set, HSAR-DETR achieves 50.8% $mA{P}_{50}$ and 31.4% $mA{P}_{50:95}$, outperforming the RT-DETR baseline by 4.2 and 3.0 percentage points, respectively. Full article

The construction industry remains among the most dangerous occupations in the world and safety management in this industry remains hierarchical in nature, being designed by engineers, regulatory agencies, and site managers and not by those who actually face the highest risk of danger. This paper argues that this asymmetry in who gets to design safety is systemic, arising from the intersection of organisational hierarchy, national power distance, and multi-tier subcontracting, all of which erode both accountability and worker voice. Drawing on a theoretical framework that integrates the safety-voice, power-distance, and systems-theoretic accident model and processes (STAMP) literatures with a structural analysis of Istanbul’s construction sector, the paper shows how legal, contractual, and cultural mechanisms suppress the feedback processes that resilience-oriented safety management depends on. Drawing on Turkish labour-law thresholds for union recognition, subcontracting data, and a case of worker protest during the construction of one of Istanbul’s largest infrastructure projects, the paper identifies three mechanisms of exclusion: statutory representation thresholds that structurally exclude construction workers, subcontracting chains that diffuse responsibility for safety outcomes away from the workers who bear the risk, and organisational hierarchies that discourage safety voice long before it reaches any formal channel. The paper concludes with a set of design principles for reintegrating worker feedback into safety control structures, offered as a contribution to resilience engineering practice. Full article

Central heating systems often rely on uniform regulation, making it difficult to meet the differentiated comfort demands of users and frequently leading to overheating. This paper proposes an end-to-end, data-driven framework that systematically couples density-adaptive clustering with dynamic response modeling for precision heating control. First, an adaptive DBSCAN (Density-Based Spatial Clustering of Applications with Noise) algorithm is developed, which automatically determines its parameters via k-distance graph initialization, differential evolution optimization, and hierarchical clustering post-processing. Without requiring a pre-set cluster number, it consistently identifies four typical daily room temperature fluctuation patterns. Validated on 120-day data from a residential community in Luoyang, the first four clusters cover over 80% of users, and the clustering quality approaches that of manually optimized conventional methods. Second, multi-input ARX (Autoregressive with Exogenous Inputs) models are built for the representative user of each cluster to characterize dynamic responses to supply water temperature, flow rate, and outdoor temperature. Rolling prediction for the entire community achieves an RMSE of 0.24 °C and an R² of 0.93. Finally, a differentiated regulation strategy combining main-cluster supply temperature control and small-cluster flow compensation is designed. Simulation results demonstrate that this strategy drives the room temperatures of all clusters significantly toward the 20 °C comfort target, with a marked reduction in standard deviation. The primary contribution of this study lies in the construction of a reproducible, closed-loop pipeline—from raw room temperature data to demand-based regulation logic—offering a quantitative basis for central heating systems transitioning from passive balancing to data-driven, classified control. Full article

A prescribed time-adaptive event-triggered sliding mode control method is proposed for a second-order aircraft antiskid braking system with unmeasurable longitudinal velocity, subject to unknown actuator faults and external disturbances. Based on the time scale transformation technique, a prescribed-time observer is constructed to estimate the unmeasurable longitudinal velocity. A practical prescribed-time super-twisting observer with a saturated gain is designed to estimate the disturbance. Within the prescribed time convergence framework, an adaptive update law and a nonsingular integral sliding surface are developed to compensate for actuator faults. Building on this, a time-varying dynamic threshold event-triggering mechanism is incorporated into the prescribed time-sliding mode control process, while excluding Zeno behavior and reducing the control update frequency. The aforementioned prescribed-time observers and the event-triggered adaptive sliding mode controller form a strict temporal hierarchical architecture. Based on Lyapunov stability theory, it is proved that the closed-loop system is practically prescribed-time stable and that all closed-loop signals are uniformly ultimately bounded. Comparative simulation results verify the effectiveness of the proposed method. Full article

Fruit and vegetable supply chains generate heterogeneous data across production, storage, logistics, and sales, creating challenges for trusted data sharing, privacy protection, and real-time traceability across distributed supply-chain information systems. Conventional single-chain blockchains suffer from limited scalability, data redundancy, and low retrieval efficiency, making them inadequate for high-frequency full-process information management. This study proposes a multi-chain blockchain framework for trusted full-process information management of fruit and vegetable supply chains. The framework integrates traceability, enterprise, notary, and regulatory chains to support hierarchical data management and privacy isolation. A reputation-based notary node election mechanism and a threshold-signature scheme based on Shamir secret sharing are designed to enhance cross-chain security and distributed regulatory consensus. To improve retrieval efficiency, a Cuckoo-Augmented Merkle Tree (CMerkle) and a skip-list-based block index are developed. Simulation results show that all malicious nodes were restricted by the 19th round, signature aggregation required 70.16 ms in a 500-node setting, and CMerkle achieved retrieval speedups of 14.7 and 153 times at data scales of 500 and 10,000 records, respectively. The framework supports trusted data governance, real-time traceability, privacy-preserving sharing, and regulatory decision support in blockchain-enabled supply-chain information systems. Full article

Photovoltaic converters operate under varying climatic conditions and non-ideal grid regimes, but factor importance is often assessed either through isolated local metrics or through pooled operating data that hide regime shifts and interaction effects. This study develops a hierarchical framework for sensitivity analysis of operating profiles of grid-connected PV converters under climatic and grid uncertainty. A compact operating-profile formulation is introduced that relates solar radiation, cell and ambient temperature, grid voltage, load, and selected design/control parameters to active power, efficiency, power factor, harmonic distortion, DC bus ripple, clipping behavior, and thermal headroom. The proposed workflow combines local normalized sensitivities for fast ranking around nominal conditions, Morris screening for factor reduction, and Sobol/Saltelli variance-based indices for global prioritization under uncertainty. The framework is demonstrated on a 100 kW synthetic reduced-order benchmark representing a three-phase two-level grid-connected PV inverter with an LCL filter. To clarify the scope of validity, the reduced-order model is cross-checked against switching-level simulations for representative nominal, clipping-prone, high-temperature and grid-stress operating windows. The results show that factor importance is not universal, but depends on the selected KPI, operating regime and uncertainty scenario. In the considered benchmark, grid voltage, cell temperature and equivalent thermal resistance are the dominant total-effect contributors, while the strongest second-order contribution appears between grid voltage and filter inductance under grid-stress conditions. The proposed framework is therefore intended as a reproducible, regime-aware sensitivity workflow rather than as a universal ranking of PV converter parameters. Full article

Overlapping optical cells create geometry-dependent inter-cell interference, while receiver-geometry and channel-estimation errors can reverse the effective non-orthogonal multiple access (NOMA) decoding order and increase successive interference cancelation (SIC) failures. This paper develops a chance-constrained receiver–scheduler co-design framework for a multi-cell NOMA visible-light communication network with an asymmetrically clipped DC-biased optical orthogonal frequency-division multiplexing physical layer. Correlated position, photodetector-orientation, and channel-estimation errors are propagated through nonlinear geometry-based scenarios. For each SIC direction, a joint three-SINR event defines a layer-, resource-, and direction-labeled probabilistic decodability graph. Candidate NOMA and orthogonal modes are screened on optimization scenarios, admitted by independent one-sided confidence bounds, and selected through resource-constrained mixed-integer linear programming. With the matching fixed, hierarchical powers are adapted under empirical conditional-value-at-risk constraints using trust-region sequential quadratic programming. Because candidate-edge certificates need not remain valid after global matching and power redistribution, the frozen complete assignment is independently recertified before held-out testing. Under the specified uncertainty generator, the proposed method maintains selected-pair outage probabilities of approximately $2.7\times {10}^{-3}$–$3.3\times {10}^{-3}$ over the half-power-angle sweep, compared with $0.027$–$0.060$ for nominal-CSI allocation. Additional experiments quantify network-wide outage, model misspecification, unbalanced deployments, feasibility, and computational cost. The results support reliable slow-timescale scheduling under the adopted link and uncertainty models, without implying distribution-free, waveform-level, or real-time guarantees. Full article

With the rapid development of Industrial Internet of Things (IIoT), large amounts of sensor data generated by industrial devices and edge nodes have become the basis of intelligent manufacturing applications. Collaborative modeling on these distributed data is important for tasks such as anomaly detection, equipment monitoring, and predictive maintenance. Federated learning offers a practical way to train models without exposing raw sensor data, but it still faces privacy leakage and malicious poisoning attacks. To address these issues, this paper proposes a hierarchical privacy protection and poisoning-robust defense framework for industrial federated learning. Starting from the sensitivity differences among parameters at different model layers, the proposed method designs a hierarchical privacy-budget allocation strategy that enhances protection for sensitive information while minimizing the performance impact of perturbation. Meanwhile, a multi-layer, multi-feature anomaly-detection mechanism is adopted to identify malicious updates by jointly exploiting directional consistency, scale stability, and inter-layer similarity, and majority voting together with update clipping is used to further improve system robustness. Experiments on Fashion-MNIST, MVTec AD, and C-MAPSS demonstrate that the proposed method can effectively suppress global-model degradation under multiple poisoning attacks and achieves a favorable balance among privacy protection strength, robustness, and training efficiency. Full article

Object detection in remote sensing imagery remains challenging due to vast scale variations, complex backgrounds, and the prevalence of small, densely packed targets. Existing CNN-based detectors are limited by restricted receptive fields, while Transformer-based methods incur prohibitive computational overhead for high-resolution inputs. In this paper, we propose SFSMamba-DETR, a detection framework that integrates state space models with Dual-Scale Window Attention for efficient and accurate remote sensing object detection. Specifically, we design a Selective Feature Scanning (SFS) module that uses the Mamba-based 2D Selective Scan mechanism to model long-range spatial dependencies with linear computational complexity. To capture both fine-grained local patterns and broader contextual cues simultaneously, we introduce a Dual-Scale Window Attention (DSWA) mechanism that operates at two complementary window scales with multi-kernel convolution bridging. These modules are orchestrated within a Cross-scale Feature Aggregation Module (CFAM) that performs hierarchical multi-scale fusion in a hybrid encoder. Extensive experiments on three primary benchmarks (MAR20, UCAS-AOD, and the Jilin-1 Satellite Aircraft Detection Dataset), together with supplementary results on DOTA and DIOR, demonstrate that SFSMamba-DETR achieves strong detection accuracy while maintaining competitive inference speed. Full article

The Grey Wolf Optimizer (GWO) is a popular metaheuristic, yet it often suffers from premature convergence and rapid diversity loss in complex, high-dimensional, or highly constrained optimization problems. This paper introduces HAGWO, a novel Hierarchical Adversarial Grey Wolf Optimizer that addresses these limitations through three synergistic enhancements: dynamic hierarchical population stratification, adaptive Levy flight perturbation, and hierarchical adversarial-like position updating. These mechanisms enable adaptive balancing of global exploration and local exploitation while preserving population diversity throughout the search process. Extensive experiments on the CEC 2017 bound-constrained benchmark suite across 30D, 50D, and 100D dimensions demonstrate that HAGWO achieves superior overall performance among eight state-of-the-art algorithms, with statistically significant advantages confirmed by Friedman mean ranks and Wilcoxon signed-rank tests. When adapted to the strongly NP-hard three-dimensional bin packing problem with identical bins (3D-SBSBPP), HAGWO delivers highly competitive results, outperforming the well-established BRKGA and most other metaheuristics while closely approaching the original GWO in solution quality and exhibiting exceptional run-to-run stability. By rigorously evaluating HAGWO across both high-dimensional continuous benchmarks and a practical constrained combinatorial application, this study validates the effectiveness of its hierarchical adversarial-like framework and provides valuable insights into algorithm design and transferability across different problem domains. Full article

This study evaluates the unique ecological and floristic identity of the coastal dune vegetation in the Enez delta and adjacent dune coast of Saros Bay (southern Thrace) by comparing it with diverse dune systems across the Anatolian peninsula. A comprehensive data matrix of 97 phytosociological relevés across nine representative coastal dunes spanning the East Mediterranean, Aegean, Marmara, and Black Sea coasts was analyzed. Methodologically, univariate non-parametric approaches (Friedman variance analysis and Durbin–Conover tests) were integrated with multivariate techniques, including Hierarchical Cluster Analysis and Principal Coordinates Analysis (PCoA) based on a Bray–Curtis dissimilarity matrix. Ephedra distachya ssp. monostachya was found to be the most characteristic and differentiating taxa from the clustering. Friedman test results demonstrated highly heterogeneous species abundance across localities (χ² = 27.2, p < 0.001). The multivariate synthesis revealed a profound ecological decoupling for Saros Bay dunes: while macroclimatic filtering forces a powerful functional convergence with arid Mediterranean models dominated by therophyte, strict composition-based metrics isolate Saros Bay coastal dunes into an entirely independent taxonomic clade. PCoA ordination confirmed this distinctiveness (p < 0.05 against six of the eight national localities), with the first two axes explaining 33.58% of the total variation (Axis 1: 19.66%, Axis 2: 13.92%). This isolation is driven by a high density of Irano-Turanian elements and specialized local lineages like Silene kotschyi. Conversely, a sharp latitudinal bio-climatic macro-gradient was mapped, showing a transition toward humid Black Sea systems strictly dictated by macroclimatic filtering rather than biotic competition (p > 0.05). To preserve these specialized niches, designating coastal dunes of Saros Bay as a Priority Conservation Unit and establishing international, transboundary catchment to coast monitoring frameworks are essential. Full article

LiDAR point clouds are sparse, irregular, and unevenly distributed, which makes representative feature extraction challenging for 3D object detection. Currently, Mamba modules have been increasingly applied to 3D object detection due to their ability to efficiently model global spatial dependencies. However, preserving geometric structures while modeling global spatial dependencies remains challenging in Mamba-based frameworks. In view of this, this paper proposes a single-stage 3D object detection framework for LiDAR-based point clouds. First, this paper designs a hierarchical multiscale structure called Multiscale Voxel–Point Alternating Fusion (MVPF) Module. Within this module, the 2DMamba module is introduced into the feature extraction stage at each scale to model global spatial dependencies in the BEV plane through selective state-space scanning. Second, we design a Z-to-Channel cross-dimensional reorganization strategy that merges the voxel Z dimension into the channel dimension, yielding a BEV-form feature representation suitable for 2DMamba processing. Third, this paper proposes a Local Voxel Feature Enhancement (LVFE) module composed of a Point-to-Voxel Feature Aggregation (PVFA) module, a Voxel Densification Module (VDM), and a coordinate-indexed Voxel-to-Point Mapping (VPM) module, which enhances local voxel representations while maintaining point–voxel spatial correspondence. On the KITTI validation set, Mamba-BEV achieved higher detection accuracy and inference speed than the baseline. Under the Moderate difficulty level of AP_R11, the 3D AP for Car, Pedestrian, and Cyclist improved by 0.6%, 1.5%, and 0.4%, respectively, compared to the baseline, while the inference speed increased from 44.05 FPS to 67.11 FPS. These results demonstrate the effectiveness of the proposed Mamba-BEV for point cloud-based 3D object detection. Full article

Cross-organization cyber defense must reconcile collaborative learning with privacy and adversarial robustness, yet standard federated learning ships full gradient tensors, leaking sensitive posture and inviting Byzantine manipulation. We present FedMARL-LTI, a federated multi-agent reinforcement learning framework whose architecture answers both pressures with a single decision: each organization’s threat intelligence is shared only as a differentially private 768-dimensional semantic embedding, never as raw data. In the evaluated system, a Weight-DP-protected model-weight delta is also exchanged through the federated aggregator (the semantic abstraction embedding is a parallel channel); the privacy guarantee below is stated for the semantic abstraction channel, and an embeddings-only architecture—which the guarantee enables—is the design this points toward. The contribution is fourfold. (1) Semantic Abstraction (SA) channel: per organization, each round, the local gradient is summarized by an LLM, projected to a 768-dim embedding, L2-clipped, and Gaussian-noised before any numeric quantity leaves the host. The bottleneck reduces the aggregate noise magnitude—the expected L2 norm of the DP noise vector—from $O\left(\sqrt{d_{\mathrm{model}}}\right)$ to $O\left(\sqrt{m}\right)$ with $m=768\ll d_{\mathrm{model}}\approx 3\times {10}^5$. (2) Formal privacy analysis: the SA + DP cascade satisfies $(\epsilon ,\delta )$-DP and bounds per-round mutual information leakage by $\min \{T_{\mathrm{tok}}{\log }_{2}V,m/2{\log }_2(1+C^2/\left(m{\sigma }^2\right)\left)\right\}$, with Rényi composition over $T$ federation rounds. Scope of the guarantee: this bound certifies (i) the semantic-abstraction channel. It does not by itself cover (ii) the weight-aggregation channel, whose Weight-DP protection is analyzed separately, nor (iii) the whole deployed system, which is the composition of the two. We therefore state the ≈1.4-bit/MI bound as a per-round guarantee on information leaving the organization through the SA channel not over every byte the system emits; an embeddings-only configuration—which this bound enables—closes the gap to a whole-system guarantee. (3) Byzantine-resilient ClippedClustering aggregator combining L2 clipping with cosine-similarity clustering. (4) Hierarchical MARL policy with threat-profile-aware LLM-IRR reward shaping, wired end-to-end and disclosed honestly (the evaluated system uses a deterministic Johnson–Lindenstrauss projection in place of the LLM call for reproducibility; the architecture is thus LLM-compatible rather than dependent on a specific model, and a full LLM deployment is the planned extension). We evaluate on CybORG CAGE-4 with $n=5$ organizations, 30 federation rounds × 5 episodes × 100 steps per round. Releasing the SA channel in parallel shows no statistically detectable reward cost at N = 5 vs. the no-privacy baseline; this is measured at reward-shaping coefficient β = 0, so it establishes that the private semantic release does not disturb weight-channel training rather than that semantic sharing improves defense: SA-only $\Delta$reward = $+4.58$ ($t=+1.37$, NS), dual SA + Weight-DP $\Delta$reward = $+4.31$ ($t=+1.30$, NS), all $N=5$ seeds, all $\left|t\right|<1.4$. A controlled signal/noise probe confirms a $19.58\times$ improvement of SA over Weight-DP at a fixed DP budget—matching the predicted $\sqrt{d/m}\approx 19.8$. Under Byzantine sign_flip at 30% ($N=15$), ClippedClustering is directionally strongest ($F_1=0.025$ vs. FedAvg $0.020$, Krum $0.016$) but the edge is not statistically significant (CC vs. Krum $t=+1.59$, $p=0.15$, $d=+0.58$; the earlier $N=5$ “$3.4\times$” gap was small-sample optimism); its Byzantine behavior is on the harsher random_noise attack. Under a corrected implementation, the undefended baselines do not diverge or collapse; the earlier reading (Krum $0.002$, ClippedClustering $0.020$) was a noise-injection artifact and is withdrawn; ClippedClustering is now directionally best on F1 but not significantly, and trails Krum on reward (superseded Cohen’s $d=+3.77$). The cooperative-PPO family (MAPPO, IPPO) outperforms value/actor-critic (QMIX, MADDPG) by $\approx 20$ reward units, $p<0.001$. All host-level $F_1$ values stay below $0.05$ at the 15K-step training horizon used here; the relative claims of the paper (no detectable privacy reward cost, ClippedClustering’s competitive (not decisive) Byzantine behavior on the harsher attacks, cooperative-PPO dominance) are unaffected by this scope. A 200K-step long-horizon replication lifts $F_1$ above the 15K plateau (to $\approx 0.044$, $N=5$)—confirming that horizon, not the privacy/Byzantine machinery, gates absolute accuracy—but a finer 60-checkpoint run shows the climb is volatile and non-monotonic and does not reach deployment-grade, an honest stability-not-compute limitation. FedMARL-LTI is therefore presented as a proof-of-concept for the relative privacy and robustness trade-offs it isolates, not as an operationally deployable cyber defense system. We release all 141 raw run JSON outputs (Phases 1–3, the L4 backend comparison, and the algorithm/aggregator baselines), the figures, and analysis scripts for replication. Full article