RIM: A Retrieval-In-Matching Framework for Cross-Domain Global Visual Localization of UAVs
作者: Xin Li, Siyuan Duan, Shang Wang, Zhimin Mao, Bingliang Hu, Geng Zhang
分类: cs.CV
发布日期: 2026-07-22
备注: 56 pages, 10 figures, and 18 tables. Supplementary material is included
💡 一句话要点
提出RIM框架以解决无人机跨域视觉定位问题
🎯 匹配领域: 支柱七：动作重定向 (Motion Retargeting) 支柱九：具身大模型 (Embodied Foundation Models)
关键词: 无人机定位 视觉定位 跨域适应 深度学习 几何一致性 检索-匹配框架 DINOv2-B 局部描述符
📋 核心要点
- 现有方法在无人机视觉定位中面临跨域外观和视角变化的挑战，导致姿态估计不够鲁棒。
- 论文提出的RIM框架通过冻结DINOv2-B检索器并提炼局部描述符，优化了检索和匹配过程。
- 实验结果表明，RIM在EPFL和长安公园数据集上分别提高了Recall@1 8.55/13.77和4.45/8.94个百分点，且定位速度显著提升。
📝 摘要（中文）
无人机的全球视觉定位利用遥感参考地图受到越来越多的关注。然而，无人机与参考图像之间的获取时间和成像平台差异导致了显著的跨域外观和视角变化，挑战了六自由度（6-DoF）姿态估计的鲁棒性。本文通过从Google 3D Tiles中采样无人机视角参考视图来应对这些变化，并提出了一种两阶段的跨域微调方案，利用姿态近似正样本和地理上远离的困难负样本。同时，局部几何一致性重新排序Top-K候选项。我们进一步提出了检索-匹配（RIM）框架，冻结适应后的DINOv2-B检索器，并提炼出一个局部描述符解码器，重用其标记字段。RIM在EPFL Urbanscape和自收集的长安公园数据集上进行零-shot评估，均超越了十个近期检索基线。
🔬 方法详解
问题定义：本文旨在解决无人机在跨域视觉定位中由于获取时间和成像平台差异导致的外观和视角变化问题。现有方法在处理这些变化时表现不佳，影响了六自由度姿态估计的准确性和鲁棒性。
核心思路：论文提出的RIM框架通过从Google 3D Tiles中采样无人机视角参考视图，结合两阶段的跨域微调策略，利用姿态近似正样本和地理上远离的困难负样本来增强模型的适应性。
技术框架：RIM框架主要包括两个阶段：首先是检索阶段，使用冻结的DINOv2-B检索器进行候选项的检索；其次是匹配阶段，利用局部描述符解码器进行精细匹配和几何一致性验证。
关键创新：RIM的核心创新在于通过冻结适应后的DINOv2-B检索器，消除了对第二个基础模型的需求，同时保留了检索描述符的构造。这种设计显著提高了处理效率和准确性。
关键设计：在模型设计中，采用了局部描述符解码器与浅层VGG19细节流的结合，确保了信息的有效传递和处理。此外，模型在损失函数和参数设置上进行了优化，以适应跨域场景的需求。
🖼️ 关键图片
📊 实验亮点
实验结果显示，RIM在EPFL Urbanscape和长安公园数据集上分别提高了Recall@1 8.55/13.77和4.45/8.94个百分点，且在Top-K=5的完整定位查询中，端到端处理时间为67.9毫秒，比最强的稀疏匹配基线快1.8倍，且比RoMa快40倍，表现出色。
🎯 应用场景
该研究的潜在应用领域包括无人机在城市环境中的导航、灾后评估、环境监测等场景。通过提高无人机在GNSS受限环境下的定位能力，RIM框架能够为无人机的自主飞行和任务执行提供更可靠的支持，具有重要的实际价值和未来影响。
📄 摘要（原文）
Global visual localization of unmanned aerial vehicles (UAVs) using remote-sensing reference maps has attracted increasing attention. However, acquisition-time and imaging-platform differences between UAV and reference imagery induce substantial cross-domain appearance and viewpoint shifts, challenging robust six-degree-of-freedom (6-DoF) pose estimation. We address these shifts by sampling UAV-viewpoint reference views from Google 3D Tiles across locations, altitudes, and orientations. A two-stage cross-domain fine-tuning recipe adapts SALAD using pose-near positives and geographically distant hard negatives, while local geometric consistency re-ranks the Top-K candidates. We further propose Retrieval-In-Matching (RIM), which freezes the adapted DINOv2-B retriever and distils a local-descriptor decoder that reuses its token field alongside a shallow VGG19 detail stream. One query-side DINOv2-B forward thus serves both SALAD retrieval and local description, eliminating a second foundation-model backbone while preserving retrieval descriptors by construction. We evaluate RIM zero-shot on the reconstructed EPFL Urbanscape and self-collected Chang'an Park datasets, both geographically disjoint from the training data. RIM outperforms ten recent retrieval baseline families. At 25/50 m under the full 3D distance metric, it improves Recall@1 over SALAD by 8.55/13.77 percentage points on EPFL and 4.45/8.94 points on Park. At Top-K=5, the complete measured localization query, including retrieval, candidate matching, and robust geometric verification, takes 67.9 ms end-to-end: 1.8 times faster than the strongest separate sparse-matching baseline and over 40 times faster than RoMa, while achieving comparable re-ranking accuracy. These results establish an efficient and deployable pipeline for UAV global visual localization in GNSS-challenged environments.