10 July 2025 in IEEE Robotics and Automation Letters
Visual geo-localization empowers unmanned aerial vehicles (UAVs) to navigate in Global Navigation Satellite System (GNSS)-denied environments. This capability is typically formulated as an image retrieval task, where UAV-acquired images are matched against geo-tagged satellite image patches. However, existing datasets are predominantly limited to single urban scenarios and often assume an ideal spatial alignment between UAV and satellite, which is unrealistic in real-world settings. To address these limitations, we present World-UAV, a large-vocabulary dataset that offers: (1) extensive scene diversity, spanning 27 unique geographical categories, and (2) realistic spatial discrepancies, incorporating significant geometric variations in rotation and scale between UAV and satellite image pairs. These geometric transformations pose substantial challenges to current global descriptor-based methods, which exhibit marked performance degradation in such scenarios. We propose UAVPlace, a novel learning approach incorporating geometric transformation encoding modules to integrate multi-perspective transformation features, thereby generating transformation-invariant global descriptors. Extensive experiments demonstrate the effectiveness of our method.
6 October 2025 in IEEE Transactions on Neural Networks and Learning Systems
Federated learning (FL) facilitates collaborative training among multiple clients while preserving data privacy by eliminating raw data transmission. However, the inherent data heterogeneity among participants induces bias during collaborative learning, significantly degrading the performance of local models. Existing FL solutions face critical challenges in achieving efficient knowledge transmission, particularly with respect to insufficient information extraction or excessive communication costs, which result in slow convergence and inferior performance. To address these limitations, we propose a novel FL framework in a synergy of multi-level prototype-based contrastive learning (CL) and soft label generation, named FedMPS. The proposed method first constructs multi-level prototypes from different layers of the model to capture semantic information in high-level features and detailed information in low-level features. These prototypes are then utilized through CL to enhance intra-class discriminability and intra-class consistency in the feature space. In addition, a prototype-guided soft label generation module is introduced to model latent interclass relationships in the output space. Instead of exchanging model parameters, FedMPS transmits only prototypes and soft labels, effectively reducing global knowledge shift and communication costs. Extensive experimental studies on six publicly available datasets validate the effectiveness of the proposed method when compared to the current state-of-the-art FL approaches. The code is available at github.com/wenxinyang1026/FedMPS
12 January 2026 in IEEE Transactions on Geoscience and Remote Sensing
Acquiring the temperatures of a 3D city model at different times and under various climatic conditions is crucial for building energy management, diagnosis, and heat distribution analysis, contributing to sustainable city development. While 3D thermal model reconstruction has been utilized for this purpose, previous methods have faced challenges with scalability and accuracy. In this paper, we propose a novel framework for 3D thermal city reconstruction with thermal mapping guided by an RGB model. Our framework consists of four steps: (1) building a high-quality reference RGB model, (2) registering thermal images to the RGB model via view synthesis and cross-modality thermal-to-visible image matching, (3) establishing pixel-level correspondence between the RGB and thermal images, and generating thermal mapping images with pixel-level alignment, and (4) converting the thermal values on the images to temperatures, and applying texture mapping to reconstruct the high-precision 3D city temperature model. The advantage of our framework is the need to build an RGB model only once, which can then guide thermal model reconstruction at any time. Our method achieves comparable geometric accuracy to the RGB model, with a C2M distance of the order of 10e-8, and superior texture quality to the direct use of thermal images, with more than 10% improvement in PSNR and SSIM, and the best performance in LPIPS.
16 January 2026 in IEEE Robotics and Automation Letters
Visual geo-localization without GPS prior remains a significant challenge for UAV navigation. Traditional retrieval-based methods suffer from scale and rotation variances between UAV images and satellite maps, and their inference speed degrades with increasing map size. To address these challenges, we propose NGC-GeoLoc, a novel geo-coordinate regression framework designed specifically for UAV localization using satellite maps only. Our core idea is to replace retrieval by directly regressing map-space coordinates for every query pixel. Technically, our pipeline combines a visual encoder with regression head, trained with dense supervision generated from multi-scale, multi-rotation satellite crops. During inference, our method efficiently recovers the UAV pose by solving a homography using predicted dense correspondences and refining with dense feature matching. This regression-based framework enables UAV localization that is robust to rotation and scale variations with sub-second inference time, regardless of map size. In experiments on UAV-VisLoc across seven different scenes, NGC-GeoLoc surpasses the baseline in recall on seven scenes and achieves lower inference time across the board.
13 August 2026 in Pattern Recognition
13 March 2026 in IEEE Robotics and Automation Letters
Image matching serves as a core component for UAV localization guided by satellite imagery. However, this task remains highly challenging due to the extreme viewpoint discrepancies between low-altitude UAV images and nadir-view satellite maps. Existing datasets primarily focus on ground-level or high-altitude UAV imagery, lacking sufficient coverage of the geometric variations typical of real aerial scenarios. To address these limitations, we introduce AerialExtreMatch, a large-scale, high-fidelity dataset tailored for extreme-view image matching and UAV localization. It consists of approximately 1.5 million synthetic image pairs rendered from high-quality 3D scenes, simulating diverse UAV and satellite viewpoints to enable robust training of image matching models. To support fine-grained evaluation, we construct a hierarchical benchmark with 32 difficulty levels. These are defined using three geometric criteria: overlap ratio, scale variation, and pitch difference. In addition, we collect a real-world UAV localization dataset with geo-aligned reference maps of varying visual quality. Extensive experiments involving 16 representative detector-based and detector-free methods demonstrate that models trained on AerialExtreMatch achieve substantial performance gains in both image matching and real-world localization under extreme-view conditions.