Cross-view geo-localization (CVGL), which involves matching and retrieving satellite images to determine the geographic location of a ground image, is crucial in GNSS-constrained scenarios. However, this task faces significant challenges due to substantial viewpoint discrepancies, the complexity of localization scenarios, and the need for global localization. To address these issues, we propose a novel CVGL framework that integrates the vision foundational model DINOv2 with an advanced feature mixer. Our framework introduces the symmetric InfoNCE loss and incorporates near-neighbor sampling and dynamic similarity sampling strategies, significantly enhancing localization accuracy. Experimental results show that our framework surpasses existing methods across multiple public and self-built datasets. To further improve global-scale performance, we have developed CV-Cities, a novel dataset for global CVGL. CV-Cities includes 223 736 ground-satellite image pairs with geolocation data, spanning sixteen cities across six continents and covering a wide range of complex scenarios, providing a challenging benchmark for CVGL. The framework trained with CV-Cities demonstrates high localization accuracy in various test cities, highlighting its strong globalization and generalization capabilities.

Cross-view image geo-localization refers to determining the geographic location of a given query image using an image database with known location information. Existing methods mainly focus on learning discriminative image representations to optimize the distance of image feature representations in feature space without fully considering the positional relation information of the features and the information redundancy in the features themselves. Therefore, we proposed a cross-view image localization method that combines global spatial relation attention with feature aggregation. First, we utilize the lightweight global spatial relation attention to learn the spatial location structure information of features, which fully enhances the perceptual and discriminative capabilities of the model. The proposed attention has little effect on the complexity and memory occupancy of the model and can be generalized to other image-processing tasks. In addition, we introduce the sinkhorn algorithm for locally aggregated descriptors, which represents the aggregation of local features as an optimal transport problem and selectively discards useless information during the clustering and assignment of features, thus enhancing the generalization and robustness of the descriptors. Experimental results on the public University-1652, CVACT, and CVUSA datasets validate the effectiveness and superiority of the proposed method. Our code is available at https://github.com/WenjianGan/LRFR.

Building instance extraction and recognition (BEAR) extracts and further recognizes building instances in unmanned aerial vehicle (UAV) images, holds with paramount importance in urban understanding applications. To address this challenge, we propose a unified network, BEAR-Former. Given the difficulty of building instance recognition due to the small area and multiple instances in UAV images, we developed a novel multi-view learning method, Cross-Mixer. This method constructs a cross-regional branch and an intra-regional branch to, respectively, extract the global context dependencies and local spatial structural details of buildings. In the cross-regional branch, we cleverly employed cross-attention and polar coordinate relative position encoding to learn more discriminative features. To solve the BEAR problem end to end, we designed a channel group and fusion module (CGFM) as a shared encoder. The CGFM includes a channel group encoder layer to independently extract features and a channel fusion module to dig out the complementary information for multiple tasks. Additionally, an RoI enhancement strategy was designed to improve model performance. Finally, we introduced a new metric, Recall@(K, iou), to evaluate the performance of the BEAR task. Experimental results demonstrate the effectiveness of our method.

Utilizing visual place recognition (VPR) technology to ascertain the geographical location of publicly available images is a pressing issue for real-world VPR applications. Although most current VPR methods achieve favorable results under ideal conditions, their performance in complex environments, characterized by lighting variations, seasonal changes, and occlusions caused by moving objects, is generally unsatisfactory. In this study, we utilize the DINOv2 model as the backbone network for trimming and fine-tuning to extract robust image features. We propose a novel VPR architecture called DINO-Mix, which combines a foundational vision model with feature aggregation. This architecture relies on the powerful image feature extraction capabilities of foundational vision models. We employ an MLP-Mixer-based mix module to aggregate image features, resulting in globally robust and generalizable descriptors that enable high-precision VPR. We experimentally demonstrate that the proposed DINO-Mix architecture significantly outperforms current state-of-the-art (SOTA) methods. In test sets having lighting variations, seasonal changes, and occlusions (Tokyo24/7, Nordland, SF-XL-Testv1), our proposed DINO-Mix architecture achieved Top-1 accuracy rates of 91.75%, 80.18%, and 82%, respectively. Compared with SOTA methods, our architecture exhibited an average accuracy improvement of 5.14%.

Narrow field-of-view (NFoV) image cross-view localization is limited by insufficient information in the ground scene itself and cross-view domain differences, resulting in even more limited information directly corresponding to satellite imagery of the same area. This leads to low accuracy in existing methods and relatively few related studies. To address this issue, this paper innovatively employs cube projection technology to construct an NFoV ground dataset. Compared to traditional panoramic cropping methods, this method converts cylindrical projection images into central projection images, accurately simulating camera imaging geometry and providing high-fidelity data support. Secondly, this paper introduces monocular depth estimation to reduce interference information from ground images during satellite image retrieval, thereby highlighting core localization features. Therefore, a two-branch ground feature extraction algorithm (processing the original image and enhancing the image after filtering out invalid information) is designed to mine rich ground features, supplemented by an independent satellite feature extraction branch, forming a three-branch network to improve the accuracy of cross-view localization in narrow field-of-view images. Experimental results show that on the CVUSA dataset, the TOP-1 localization accuracy with a 90-degree field of view is improved from 55.9% to 89.14%; and on the CVACT dataset, the TOP-1 localization accuracy with a narrow field of view is improved from 40.6% to 84.6%.