Search Results (48)

Multimodal medical image fusion combines anatomical and functional information from different imaging modalities. However, existing methods often struggle to preserve fine anatomical structures while incorporating complementary functional information. To address this problem, we propose TAPGFusion, an anatomy-aware multimodal medical image fusion network. It employs a Multi-Scale Encoder to capture fine local details and broad anatomical structures through parallel convolutions with different receptive fields. A Detail-Enhanced Attention Block further refines the extracted features through channel, spatial, and pixel attention. In addition, a Physiological Prior-Guided Attention Block dynamically balances anatomical and functional features using MRI-conditioned prior information and edge constraints. The main contribution of TAPGFusion is a unified framework that jointly addresses multi-scale feature representation, fine-grained feature selection, and spatially adaptive anatomical–functional fusion. Extensive experiments on three public medical imaging datasets demonstrate the effectiveness and robustness of the proposed method. TAPGFusion achieves CC values above 0.83 and SSIM values above 0.78 on the CT–MRI, PET–MRI, and SPECT–MRI fusion tasks. These results indicate that the proposed method effectively preserves anatomical structures while integrating complementary functional information. Full article

High-resolution remote sensing stereo matching is challenging due to heavy computation and the difficulty of handling textureless areas, repetitive structures, and occlusions. To tackle these issues, we design BDNet (Balancing Dual-path Network)—a stereo matching network that seeks a reasonable trade-off between accuracy and efficiency for remote sensing applications. In the feature extraction stage, BDNet adopts progressive dilation with rates 5, 4, and 3, together with a decoupled multi-scale reduction (DMSR) module, which reduces multi-scale feature channels from 320 to 32. We also introduce a strip attention module to make the network more sensitive to horizontal and vertical structures commonly seen in urban scenes. For cost volume construction, the number of correlation groups is lowered from 40 to 8, in line with the compact 32-channel feature representation. For cost aggregation, a dual-path parallel hourglass architecture is designed, which preserves fine details through a high-resolution path while capturing global context through a low-resolution path. An attention-guided fusion module adaptively integrates features from both paths, improving accuracy in challenging regions such as textureless areas and disparity discontinuities. Experiments on the US3D and WHU-Stereo datasets demonstrate that BDNet achieves the best accuracy among the selected baseline methods on US3D, with D1 errors of 16.05% on Jacksonville and 11.84% on Omaha. It requires only 1.43 M parameters and 125.43 G FLOPs, achieving a favorable balance between accuracy and efficiency. Zero-shot generalization experiments on Omaha and WHU-Stereo further suggest the model’s potential for cross-domain adaptation to different satellite sensors and urban scenes. Full article

Modern remote sensing images (RSIs) provide increasingly fine spatial detail, making pronounced scale variations and complex spatial distributions of land-cover classes more apparent and thereby increasing the difficulty of semantic segmentation. Recent remote sensing semantic segmentation methods therefore have increasingly adopted Transformer- and Mamba-based models to improve global contextual modeling. However, these models often suffer from intraclass inconsistency and interclass feature confusion, leading to fragmented object structures and imprecise boundary delineation. Accordingly, we develop SPFMamba, an architecture built around Mamba that combines semantic prompting with frequency-adaptive fusion, thereby improving global context modeling and fine-detail representation. During feature reconstruction, we propose a semantic prompt global–local Mamba (SPGLM) block to jointly model global semantic information and local spatial cues. Its parallel semantic prompt global and multidirectional local perception branches promote semantically coherent and spatially continuous feature distributions, thereby effectively preserving the structural integrity of ground objects. To further alleviate cross-level semantic discrepancies and strengthen the representation of small-scale targets, we design a high-frequency adaptive fusion module (HFAFM). It first refines high-frequency responses in shallow layers to retain small-scale object structures and boundary cues. Subsequently, deep semantic priors guide local cross-attention, allowing low-level spatial cues to be selectively integrated with deeper semantic representations. Evaluations across ISPRS Vaihingen, ISPRS Potsdam, and OpenEarthMap datasets show that SPFMamba delivers favorable segmentation accuracy with only 17.85 M parameters, while showing improved preservation of object structures and fine spatial details in qualitative comparisons. Full article

Accurate remaining useful life (RUL) prediction for aero-engines is important for condition-based maintenance and safety-oriented health management. Long short-term memory (LSTM) networks are widely used for this task, but two limitations remain important in multi-sensor degradation modeling: hidden states generated over the full window are often under-utilized, and attention mechanisms may overemphasize locally fluctuating sensor readings. This paper proposes a Multi-Statistic Disentangled LSTM (MSD-LSTM) framework for aero-engine RUL prediction. The framework first applies Savitzky–Golay filtering to smooth high-frequency signal fluctuations. A hidden-state feature extraction module then combines feature-level disentangled extraction and Global Average Pooling to use the LSTM hidden-state sequence beyond the final recurrent output. In parallel, a Multi-Statistic Pooler summarizes each input window using minimum, maximum, standard deviation, and mean statistics, and its output is fused with a self-attention branch through a static-gating mechanism. On the NASA C-MAPSS benchmark, MSD-LSTM achieves RMSE values of 10.45 and 12.33 on FD001 and FD002, respectively, and ranks first in RMSE on three of the four sub-datasets and first in SCORE on two sub-datasets among the compared recent methods. Ablation and fusion analyses show that both the hidden-state extraction and statistic-guided fusion components contribute to stable RUL prediction. Full article

Wheat is a major staple crop, and storage mold growth poses a severe threat to grain safety and quality stability. Natural mold development in stored wheat exhibits subtle, localized, and highly heterogeneous characteristics. Existing unimodal methods and global fusion approaches generally suffer from insufficient local feature sensitivity, hindering fine-grained mold severity grading. To address this limitation, we propose a Mask-Guided Fine-Grained Fusion Network, a weakly supervised framework based on local RGB–HSI fusion. This framework employs a dynamic parallel A/B experimental design to construct time-matched proxy labels via weakly supervised learning. A standardized preprocessing pipeline including single-kernel extraction, foreground segmentation, and cross-modal registration is established to resolve RGB–HSI spatial misalignment, ensuring physical-level spatial consistency of multimodal features. The model incorporates a Foreground-Aware Spectral Recalibration (FASR) module to suppress background noise, a Mask-Guided Dilated Cross-modal Local Attention (MDCLA) mechanism to establish fine-grained local mappings between RGB visual phenotypes and hyperspectral responses, and a sample-level adaptive fusion strategy to dynamically weight features by modal reliability, enhancing representation of complex samples across all mold stages. Experiments show that the Mask-Guided Fine-Grained Fusion Network achieves 0.9689 classification accuracy, 0.9698 Macro-F1 score, and 0.0593 Mean Absolute Error (MAE), significantly outperforming state-of-the-art unimodal deep models and global attention fusion baselines. This work provides a proof-of-principle framework for fine-grained non-destructive mold risk assessment in stored wheat. Full article

Effectively integrating graph topology and node attributes, while assigning nodes with both semantic similarity and structural closeness to the same community, remains a key challenge in attributed graph community detection. To address this challenge, this study proposes TVHCL-CD, a two-view hierarchical contrastive learning-driven method for community detection. The proposed method constructs an attribute view and a modularity view from the node attribute matrix and the modularity matrix, respectively, to model attribute semantics and high-order community structure priors. Structure-aware two-view representations are then learned in parallel through dual-view graph attention encoders incorporating multi-order neighborhood priors. Furthermore, a structure-enhanced Graph Transformer fusion module is designed to achieve node-level adaptive fusion of the two-view representations by introducing a learnable adjacency bias into global self-attention and a view-aware gating mechanism into the feed-forward network. To align the optimization objective with community semantics, a hierarchical contrastive learning strategy is further developed. Specifically, view-level consistency contrastive learning constructs modularity-guided augmented views to improve representation robustness, while community-level semantic contrastive learning incorporates partial ground-truth labels to enhance intra-community compactness and inter-community separation. Finally, clustering is performed on the fused representations to obtain community partitions. Experimental results on eight real-world attributed graphs and the generated tree-like attributed graph Tree-2500 indicate that TVHCL-CD achieves competitive performance under the semi-supervised transductive setting, while ablation results support the contributions of its main components. Full article

Existing single-image super-resolution methods for remote sensing images suffer from insufficient global receptive fields, weak high-frequency texture recovery, and excessive computational complexity. To address these issues, this paper proposes DFSMamba, a novel spatial–frequency collaborative modeling framework. First, Semantic Continuous-Sparse Attention enhances semantic perception through dynamic chunking and sparse connections while maintaining linear complexity, effectively alleviating the semantic truncation problem caused by fixed window partitioning. Second, the Adaptive State-Space Module employs parallel forward and backward state-space model branches to achieve bidirectional long-range dependency modeling and introduces an activation-guided feature fusion mechanism to adaptively enhance semantically relevant regions. Third, the Discrete Fourier Transform Module maps images to the frequency domain, establishes a global lossless receptive field, and explicitly enhances high-frequency details, compensating for the insufficient utilization of frequency-domain information in pure spatial-domain methods. Experiments on five public datasets demonstrate that DFSMamba outperforms mainstream CNN, Transformer, and Mamba-based methods across ×2 to ×4 scales. On the AID×3 task, it achieves a PSNR of 31.48 dB, exceeding MambaIRv2 by 1.07 dB. Ablation studies verify the positive synergistic effect of the three modules, with the full configuration achieving a PSNR improvement of 0.85 dB over the single-module setup. Fine-grained category, multi-scale input, and loss function experiments further confirm its robustness and generalization capability, particularly in edge and texture detail reconstruction. Full article

Dual 6 degrees of freedom (6-DOF) series-parallel cooperative robot actuators are core execution components in modern intelligent manufacturing systems, which are widely used in high-end manufacturing scenarios such as aerospace precision assembly, laser precision machining, and core component assembly of new energy vehicles. However, in actual manufacturing processes, the pose deviation between theoretical model prediction and actual motion execution of the actuator, caused by kinematic model mismatch, unquantified core parameters, incomplete error processing chain, and complex on-site environmental interference, severely restricts the assembly accuracy, product qualification rate and production efficiency of the manufacturing system. To address these critical pain points of robot actuators in precision manufacturing systems, this paper proposes a four-layer hierarchical decoupling multi-modal fusion calibration algorithm for high-precision pose control of dual series-parallel robot actuators. The algorithm integrates singular value decomposition (SVD) for cross-structure coordinate alignment of heterogeneous actuators, chaotic mapping-enhanced particle swarm optimization (PSO) for nonlinear error suppression of the actuator system, attention-enhanced deep residual network (DRN) for unmodeled residual learning of the actuator, and Kalman filter (KF) for dynamic noise reduction in the manufacturing process. Meanwhile, a full-chain error transfer model of the actuator system in the manufacturing process is constructed, and the core parameters of the algorithm are quantified via dimensional sensitivity analysis and orthogonal experiments. Experimental results show that the static position error of the actuator system after calibration reaches 1.4 ± 0.08 mm, and the static pose error reaches 0.0059 ± 0.0003 rad in the laboratory environment; in the engineering application of laser precision machining in an actual manufacturing line, the position error and pose error only increase by 8.6% and 6.8% respectively, maintaining high stability in industrial manufacturing scenarios. Compared with mainstream calibration methods, the proposed algorithm reduces the position error and pose error of the actuator by up to 55.7% and 17.9% respectively, with lower computational complexity and higher engineering reproducibility. This work constructs an end-to-end error suppression chain with quantitative parameter criteria for the series-parallel actuator system in manufacturing systems, which provides a reliable high-precision calibration solution for industrial dual-robot cooperative manufacturing and has important guiding significance for improving the motion accuracy and operation stability of actuators in precision manufacturing systems. Full article

High-resolution range profile (HRRP) target recognition has advanced with deep learning, yet most existing methods rely primarily on data-driven feature extraction and still face challenges in physical interpretability and noise robustness. To address these issues, this paper proposes a physics-guided dual-branch fusion (PGDBF) model. It consists of two parallel branches: a data-driven branch extracts discriminative features from raw HRRPs, while a physics-guided branch estimates sparse peak parameters (position and intensity) and reconstructs the signal envelope under sparsity constraints. Cross-attention adaptively fuses the two branches. Experiments on measured ten-class aircraft HRRP data show that PGDBF achieves higher accuracy and improved robustness under additive Gaussian noise in the evaluated fixed-route scenario. Visualizations confirm that the estimated peaks align with dominant HRRP energy, linking model variables to physically meaningful peak locations and intensities. These results suggest that integrating explicit peak parameter estimation with data-driven learning is a promising direction for improving HRRP recognition robustness and interpretability under low-SNR conditions. Full article

To address the challenge of fragmented, unstructured knowledge in the cultural relic domain, where existing entity and relation extraction models suffer from boundary confusion and feature degradation for long entities and overlapping triples, this paper proposes a subject-guided two-stage joint entity and relation extraction model tailored to cultural relic texts, and constructs a Cultural Relic Knowledge Graph System. Building on CasRel’s cascaded labeling framework, we design a Multi-Head Self-Attention Decoder Enhanced by Relative Position Encoding (MHSA-RPE) to explicitly model inter-entity positional relations and alleviate boundary confusion. We further propose a Boundary–Global Dual-Branch Subject Fusion Module (BGDSFM) to encode local boundary and global contextual features in parallel, alleviating feature degradation from simple average pooling. Experiments on DuIE2.0 and our self-built Palace Museum Cultural Relic Entity–Relation Dataset (PM-CRER) show that the proposed model achieves F1-scores of 79.4% and 75.9% respectively. It outperforms mainstream baselines, surpassing its prototype CasRel by 3.6 percentage points on PM-CRER and the latest cascaded state-of-the-art CECRel by 2.6 percentage points on DuIE2.0. Based on this model, a Chinese Cultural Relic Knowledge Graph System supporting the multimodal display of cultural relic images is constructed, providing technical references for the digital protection, dissemination and utilization of cultural relic knowledge. Full article

As the role of remote sensing continues to grow, the fusion technology of low-spatial-resolution hyperspectral images and high-spatial-resolution multispectral images has become increasingly critical. Traditional methods rely on fixed rules and exhibit poor robustness, whereas deep learning methods struggle to establish efficient interactions between local and global information due to the complexity of their underlying networks. Therefore, we propose a deep learning fusion module that combines pixel-wise adaptive wavelet transform with a spectral–spatial dual-branch extraction. Firstly, by utilizing the unique properties of the wavelet transform, it is possible to effectively preserve spectral information and extract spatial edge features, thereby achieving preliminary fusion by leveraging both low-frequency and high-frequency components. To compensate for the lack of nonlinear expression capability in the wavelet transform, a dual-branch parallel extraction of spectral and spatial features is subsequently performed in the deep learning module. The Multi-Scale Group Convolution module (MSGC) is utilized to extract spectral information, while the Spectral Compression and Spatially Guided Gating Module (SCSGM) is employed to extract spatial information, thereby enhancing the data’s adaptive capability. A bidirectional attention mechanism is interspersed within the module to capture complementary information across different scales, ultimately reconstructing a high-resolution hyperspectral image. Finally, the proposed fusion strategy demonstrates superior performance in practical image reconstruction, outperforming more than ten state-of-the-art fusion methods. Full article

Transformer-based speech enhancement (SE) architectures suffer from high computational complexity, while existing lightweight state space model (SSM) approaches are constrained to fixed one-dimensional scanning that cannot fully exploit the two-dimensional time–frequency structure of speech spectrograms. To address these limitations, we propose DOM-MUSE, a lightweight U-Net-style SE framework built upon the Mamba-2 SSM with four targeted innovations. First, a Deformable Feature Extractor (DFE) predicts per location spatial offsets that warp the feature sampling grid to align with speech formant trajectories and harmonic structures, providing geometrically coherent inputs to the state space model. Second, a DOM Mamba Block with Cross-Dimensional Gated Fusion (CDGF) deploys two parallel Mamba-2 instances scanning the time and frequency axes independently, and uses Taylor Channel Attention (TCA) to derive semantic gates that modulate each SSM output before fusion. Third, a Phase-Guided Feature Conditioner (PGFC) computes local phase-gradient gates that suppress noise-dominated activations prior to the SSM stage, making the feature extraction pathway implicitly phase-aware. Fourth, an Attention-Based Skip Connection (ABSC) replaces conventional concatenation skip connections with a learned channel gate, adaptively controlling the information flow from the encoder to the decoder. Experiments on the VoiceBank-DEMAND benchmark demonstrate that DOM-MUSE outperforms the reproduced MUSE baseline on all five evaluation metrics—including PESQ (+0.077), CSIG (+0.058), CBAK (+0.026), COVL (+0.070), and STOI (+0.002)—while reducing the parameter count by 24% (0.51 M to 0.39 M). Notably, DOM-MUSE also surpasses MUSE++ on perceptual quality metrics (PESQ +0.061, COVL +0.032) despite MUSE++ employing dynamic SNR augmentation and an augmented multi-objective loss that DOM-MUSE deliberately omits, demonstrating that the proposed architectural innovations yield genuine improvements independent of training strategy. When DOM-MUSE is additionally trained under the same augmented protocol as MUSE++, it achieves PESQ of 3.46 and COVL of 4.22, further confirming the complementary nature of architectural and training improvements. Full article

Small-object detection in unmanned aerial vehicle (UAV) images poses significant challenges due to limited pixel representation, complex backgrounds, and insufficient feature discriminability. While one-stage detectors like YOLO offer a favorable speed-accuracy trade-off, their performance on small objects is often hampered by conflicts between semantic and spatial information during multi-scale feature fusion in existing networks. To address this, we propose LHA-YOLO, a lightweight and high-accuracy network based on YOLO11. The network is built upon two core innovations. The first is the Lightweight Feature Extraction Module (LFEM), which employs a parallel spatial-channel attention mechanism to extract discriminative cross-dimensional features efficiently and with low computational cost. The second is the Divide-and-Conquer Propagation Path (DCPP) strategy. This strategy decouples and separately optimizes the handling of semantic and spatial information within its bidirectional propagation paths. To achieve this, the top-down path utilizes the Channel Attention-guided Semantic Aggregation (CASA) module to enhance semantic consistency. In parallel, the bottom-up path employs the Spatial Attention-guided Detail Aggregation (SADA) module to preserve spatial fidelity. Extensive evaluation on the VisDrone and UAVDT datasets shows that LHA-YOLO strikes a favorable balance between performance and efficiency. On VisDrone, it improves mAP50 from 39.4% to 41.6% and mAP50–95 from 23.5% to 24.9% over YOLOv11s. On UAVDT, it raises mAP50 from 32.2% to 36.9% and mAP50–95 from 19.4% to 22.9%, while reducing GFLOPs from 21.3 to 18.8. These results confirm the efficacy of our design for real-time UAV applications. Full article

Accurate cloud and cloud shadow segmentation is a critical prerequisite for remote sensing image preprocessing. However, this task remains challenging due to the directional continuity of projected cloud shadows, the radiometric ambiguity between low-reflectance shadows and other dark surfaces, and the difficulty of preserving semantic consistency and fine boundaries in complex scenes. To address these issues, this paper proposes a Directional Statistical Fusion Network (DSFNet) based on an enhanced DeepLabV3+ architecture. Specifically, a Directional Scale Refinement Module (DSRM) is introduced in parallel with Atrous Spatial Pyramid Pooling to strengthen the representation of direction-sensitive cloud-shadow structures and multi-scale cloud regions. An Adaptive Statistical Context Attention (ASCA) module is further designed to perform robust feature modulation by jointly exploiting global statistics, edge-aware statistics, and median-based normalization, thereby suppressing anomalous responses under heterogeneous backgrounds. In the decoder, an Adaptive Grouped Multi-scale Fusion (AGMF) module is employed to adaptively fuse shallow detail features and high-level semantic features through discrepancy-guided grouped gating, improving structural consistency and boundary recovery. In addition, a hybrid loss is adopted to further optimize segmentation. Experiments on the GF1_WHU dataset show that DSFNet achieves 76.97% mIoU, demonstrating strong effectiveness and robustness in complex remote sensing scenes. Full article

Deepfake technology poses significant threats to information authenticity and social trust, necessitating effective detection methods. However, existing detection approaches predominantly rely on high-complexity network architectures that, while accurate in controlled environments, suffer from prohibitive computational costs that hinder deployment in resource-constrained scenarios such as social media platforms. To address this efficiency-accuracy dilemma, we propose a lightweight face forgery detection method that systematically learns multi-scale forgery traces. Our approach features a four-stage lightweight architecture that hierarchically extracts features from local textures to global semantics, mimicking the human visual system. Within each stage, a multi-scale dynamic perception mechanism divides feature channels into parallel groups equipped with lightweight attention modules to capture forgery cues spanning pixel-level anomalies, local structures, regional patterns, and semantic inconsistencies. Furthermore, rather than relying on conventional feature fusion that risks suppressing subtle artifacts, we introduce a novel Context-Guided Dynamic Convolution. This mechanism uses mid-level spatial anomalies as active anchors to dynamically modulate high-level semantic filters, with the goal of mitigating the disconnect between semantic content and forgery evidence. Our model achieves strong performance, yielding an AUC of 91.98% on FaceForensics++ and 93.50% on DeepFake Detection Challenge, outperforming current state-of-the-art lightweight methods. Furthermore, compared to heavy Vision Transformers, our model achieves a superior performance-efficiency trade-off, requiring only 3.06 M parameters and 1.36 G FLOPs, making it highly suitable for real-time, resource-constrained deployment. Full article