Explore the vast range of titles available.

To achieve non-contact and accurate identification of key temperature regions in cattle, this study proposes an improved semantic segmentation network named T-UNet based on dual-modality images. The network features a dual-branch encoder to extract RGB and thermal infrared features independently. A Selective Thermal Modality Attention (STM) module is integrated at each layer for multi-scale semantic alignment, while CBAM attention and a Context Enhanced Decoder (CED) are introduced to enhance boundary reconstruction and small-object segmentation. Experimental results show that T-UNet outperforms baseline models including U-Net, DeepLabV3+, and Mask R-CNN on a custom dataset, achieving 87.36% mIoU, 93.03% mDice, and 93.03% F1-score. Ablation studies and visual comparisons further confirm the effectiveness of each module. This method provides a high-precision and efficient image processing solution for infrared-based cattle temperature monitoring, with strong application potential in smart livestock farming.

Advancements in satellite sensor technology have enabled access to diverse remote sensing (RS) data from multiple platforms. Hyperspectral Image (HSI) data offers rich spectral detail for material identification, while LiDAR captures high-resolution 3D structural information, making the two modalities naturally complementary. By fusing HSI and LiDAR, we can mitigate the limitations of each and improve tasks like land cover classification, vegetation analysis, and terrain mapping through more robust spectral–spatial feature representation. However, traditional multi-scale feature fusion models often struggle with aligning features effectively, which can lead to redundant outputs and diminished spatial clarity. To address these issues, we propose the Cross Attention Bridge for HSI and LiDAR (CAB-HL), a novel dual-path framework that employs a multi-stage cross-attention mechanism to guide the interaction between spectral and spatial features. In CAB-HL, features from each modality are refined across three progressive stages using cross-attention modules, which enhance contextual alignment while preserving the distinctive characteristics of each modality. These fused representations are subsequently integrated and passed through a lightweight classification head. Extensive experiments on three benchmark RS datasets demonstrate that CAB-HL consistently outperforms existing state-of-the-art models, confirm that CAB-HL consistently outperforms in learning deep joint representations for multimodal classification tasks.

Unmanned Aerial Vehicles (UAVs) play vital roles in traffic surveillance, disaster management, and border security, highlighting the importance of reliable infrared-visible image detection under complex illumination conditions. However, UAV-based infrared-visible detection still faces challenges in multi-scale target recognition, robustness to lighting variations, and efficient cross-modal information utilization. To address these issues, this study proposes a lightweight Dual-modality Adaptive Pyramid Transformer (DAP) module integrated into the YOLOv8 framework. The DAP module employs a hierarchical self-attention mechanism and a residual fusion structure to achieve adaptive multi-scale representation and cross-modal semantic alignment while preserving modality-specific features. This design enables effective feature fusion with reduced computational cost, enhancing detection accuracy in complex environments. Experiments on the DroneVehicle and LLVIP datasets demonstrate that the proposed DAP-based YOLOv8 achieves mAP scores of 61.2% and 62.1%, respectively, outperforming conventional methods. The results validate the capability of the DAP module to optimize cross-modal feature interaction and improve UAV real-time infrared-visible target detection, offering a practical and efficient solution for UAV applications such as traffic monitoring and disaster response.

Wind power forecasting is critical to grid stability and renewable energy integration. However, existing deep learning methods struggle to incorporate semantic domain knowledge from textual information, exhibit limited generalization with scarce training data, and require high computational costs for extensive fine-tuning. Large language models (LLMs) offer a promising solution through their semantic representations, few-shot learning capabilities, and multimodal processing abilities. This paper proposes LLM-WPFNet, a dual-modality fusion framework that integrates frozen pre-trained LLMs with time-series analysis for wind power forecasting. The key insight is encoding temporal patterns as structured textual prompts to enable semantic guidance from frozen LLMs without fine-tuning. LLM-WPFNet employs two parallel encoding branches to extract complementary features from time series and textual prompts, unified through asymmetric multi-head attention fusion that enables selective semantic knowledge transfer from frozen LLM embeddings to enhance temporal representations. By maintaining the LLM frozen, our method achieves computational efficiency while leveraging robust semantic representations. Extensive experiments on four wind farm datasets (36–200 MW) across five prediction horizons (1–24 h) demonstrate that LLM-WPFNet consistently outperforms state-of-the-art baselines by 11% in MAE and RMSE. Notably, with only 10% of training data, it achieves a 17.6% improvement over the best baseline, validating its effectiveness in both standard and data-scarce scenarios. These results highlight the effectiveness and robustness of the dual-modality fusion design in predicting wind power under complex real-world conditions.

Photovoltaic (PV) module faults, such as hotspots, diode short circuits, occlusions, and shadows, degrade power generation efficiency and safety. Existing manual inspection and single-modality methods show limited robustness under complex conditions, especially with illumination variations and weak thermal responses, while most deep learning approaches fail to exploit the complementarity of visible and infrared modalities. To address this issue, a dual-modality visible-infrared fusion framework based on YOLO11 is proposed, integrating a multi-scale pyramid pooling and dilated convolution module (MSPPD), a gradient-aware fusion module (GAFusion), and a dynamic convolution and element-wise scaling detection head (Detect-DEhead). GAFusion enhances cross-modal structural consistency and reduces feature misalignment and information loss during fusion by introducing gradient-aware feature interaction. Shape-IoU loss is employed to improve localization accuracy. The proposed method improves mean average precision (mAP)@0.5 from 86.7% to 88.1%, while reducing parameters, computational cost, and model size from 4.3 M to 3.7 M, 11.42 GFLOPs to 9.37 GFLOPs, and 9.1 MB to 7.9 MB, respectively. With Shape-IoU, mAP@0.5 reaches 88.4%, and recall increases from 78.5% to 84.9%. Experiments on the FLIR Thermal dataset achieve gains of 2.2%, 1.6%, and 2.7% in precision, recall, and mAP@0.5. The method achieves an effective trade-off between accuracy and efficiency for intelligent PV module inspection.

Aiming at the limitations of visible images in object detection, this paper proposes a dual-branch network object detection algorithm based on dual-modality fusion of visible and infrared images. Based on YOLOv7-s, the algorithm firstly introduces a spatial attention module to enhance the model’s ability of capturing key features; secondly, to resolve the problem of inconsistent object sizes, a visible multi-scale feature fusion module is proposed, meanwhile, the structure of the SimCSPSPPF module (an improved spatial pyramid pooling module) from YOLOv6 is adopted to construct an infrared multi-scale feature fusion module to efficiently extract multi-scale features from infrared images; finally, a cross-modal feature fusion module is proposed to fuse corresponding scale features from visible and infrared images. The proposed algorithm is tested on KAIST, FLIR, and GIR datasets, experimental results show that the proposed algorithm has better performance, compared with the YOLOv7-s algorithm to detect visible and infrared images separately on the KAIST dataset, the detection accuracy is improved by 18.0 and 5.1%, respectively, and detection speed is 51.8 FPS; on FLIR and GIR datasets, the proposed algorithm also demonstrates significant advantages. Furthermore, the proposed algorithm can detect objects on individual visible or infrared images while maintaining high detection accuracy.

RGB and Thermal (RGBT) tracking is an important supplement of visual object tracking for its’ unique practical and research value. However, due to the limitations of the RGBT camera, extra interferences are introduced in the data synchronously. Removing or alleviating these interferences is a crucial direction to improve the performance of the RGBT tracking task. For this problem, we propose the multi-stage matching guidance and context integration network (M2GCI) for RGBT tracking. M2GCI reorganizes the feature-encoding pipeline into two context-integrating stages that are responsible for processing primary and senior information respectively. Firstly, the primary features encoded from primary encoders are integrated by the spatially adaptive fusion strategy. Next, the senior encoders cooperating with axial external-attention redistribution strategy further extract senior features which are more reliable for target prediction. Extensive experiments on the RGBT234 and GTOT provide that the proposed M2GCI can achieve more precision and robust tracking under some difficult scenarios compared with the excellent methods.

Urban real-time rainfall-runoff prediction (URRP) is a potential nonstructural measure for urban flood management. Due to significant influence of human activities and the built environment, it is still a challenging issue to obtain satisfactory results that support effective emergency response. To obtain accurate and stable prediction results, this investigation employs rainfall and runoff datasets that mutually complementary to construct URRP model, providing more comprehensive features of runoff process. To extract features from such two kinds of datasets, a dual-modality encoder-decoder (DM-ED) model is proposed. DM-ED model employs encoder-decoder (ED) framework to enhance multi-steps prediction performance, and LSTM is embedded in the encoder and decoder layers, respectively, thus capturing high time-dependence and non-linearity of rainfall and runoff features. Then interactive dual cross (IDC) attention module is designed to capture global cross feature between rainfall and runoff. Additionally, different from common used pre-fusion approach, we propose a post-fusion (PF) module to efficiently fuse rainfall and runoff features, which can capture more comprehensive information and enhance model robustness. The DM-ED with IDC and PF model is trained and tested on urban rainfall-runoff events (January 2018 - December 2019) over a 3.52 terrain in Chongqing, China. Several experiments have been conducted on this terrain, and the experimental results show that NSE, RMSE, , MAE coefficients outperform other traditional models. The results indicate that DM-ED with IDC and PF model is expected to offer a reliable and effective method for URRP tasks.

The extensive development of image fusion techniques over the past 20 y has shown that the fusion of images from complementary modalities offers a more complete and accurate assessment of disease than do images from a single modality. Although software techniques have been successful in fusing images of the brain from different modalities, they have achieved rather limited success for other parts of the body. The recent introduction of technology that can acquire both anatomic and functional images in a single scan has addressed many of the limitations of software fusion. The combination of CT and PET was introduced commercially in 2001, followed by CT and SPECT in 2004. Clinical adoption of PET/CT has been surprisingly rapid, and despite continuing debate, the new technology has advanced the use of clinical molecular imaging, particularly for oncology.

Industrial visual inspection demands high-precision anomaly detection amid scarce annotations and unseen defects. This paper introduces a zero-shot framework leveraging multimodal feature fusion and stabilized attention pooling. CLIP’s global semantic embeddings are hierarchically aligned with DINOv2’s multi-scale structural features via a Dual-Modality Attention (DMA) mechanism, enabling effective cross-modal knowledge transfer for capturing macro- and micro-anomalies. A Stabilized Attention-based Pooling (SAP) module adaptively aggregates discriminative representations using self-generated anomaly heatmaps, enhancing localization accuracy and mitigating feature dilution. Trained solely in auxiliary datasets with multi-task segmentation and contrastive losses, the approach requires no target-domain samples. Extensive evaluation across seven benchmarks (MVTec AD, VisA, BTAD, MPDD, KSDD, DAGM, DTD-Synthetic) demonstrates state-of-the-art performance, achieving 93.4% image-level AUROC, 94.3% AP, 96.9% pixel-level AUROC, and 92.4% AUPRO on average. Ablation studies confirm the efficacy of DMA and SAP, while qualitative results highlight superior boundary precision and noise suppression. The framework offers a scalable, annotation-efficient solution for real-world industrial anomaly detection.