GeoSearcher: Anchor-Guided Progressive Reasoning for Remote Sensing Visual Grounding with Process Supervision

📄 arXiv: 2607.01050v1 📥 PDF

作者: Dianyu Wang, Yidan Zhang, Peirong Zhang, Xuyang Li, Xiaoxuan Liu, Lei Wang

分类: cs.CV

发布日期: 2026-07-01

备注: 14 pages,11 figures,7 tables

🔗 代码/项目: GITHUB


💡 一句话要点

提出GeoSearcher以解决遥感视觉定位中的小目标识别问题

🎯 匹配领域: 支柱九:具身大模型 (Embodied Foundation Models)

关键词: 遥感视觉定位 小目标识别 多模态大语言模型 渐进推理 锚点引导 过程优化 视觉线索整合

📋 核心要点

  1. 现有的遥感视觉定位方法在处理小目标和复杂查询时表现不稳定,难以有效识别目标。
  2. GeoSearcher通过锚点引导的渐进推理过程,分阶段整合视觉线索,提升了定位精度。
  3. 在多个数据集上,GeoSearcher的性能超越了当前最先进的方法,验证了其有效性。

📝 摘要(中文)

近年来,多模态大语言模型(MLLMs)在视觉定位中展现出强大的跨模态理解和坐标生成能力。然而,将这些能力迁移到遥感视觉定位(RSVG)中仍然面临挑战。高分辨率的遥感图像通常覆盖大规模场景,目标往往非常小且被许多视觉上相似的干扰物包围。同时,查询通常包含多个线索,如参考物体、空间关系和目标属性。现有基于MLLM的方法通常将RSVG视为一步坐标生成,这可能导致小目标定位和复杂查询的预测不稳定。为了解决这些挑战,本文提出了GeoSearcher,将RSVG重新定义为一种基于锚点的渐进推理过程,并通过锚点中心推理监督微调(ACR-SFT)和过程忠实的相对策略优化(PF-GRPO)两个阶段实现。实验结果表明,GeoSearcher在DIOR-RSVG、OPT-RSVG和VRS-Bench数据集上超越了现有的最先进方法。

🔬 方法详解

问题定义:本文旨在解决遥感视觉定位中小目标识别的困难,现有方法在处理复杂查询时容易出现不稳定的预测,尤其是在高分辨率图像中。

核心思路:GeoSearcher通过将RSVG重新定义为基于锚点的渐进推理过程,逐步整合位置、关系和属性线索,以提高小目标的定位精度。

技术框架:GeoSearcher的整体架构包括两个主要阶段:锚点中心推理监督微调(ACR-SFT)和过程忠实的相对策略优化(PF-GRPO)。在ACR-SFT中,模型通过锚点学习关键视觉线索,并逐步整合相关信息;在PF-GRPO中,通过过程感知奖励和推理信息样本选择器优化推理行为。

关键创新:GeoSearcher的主要创新在于其将RSVG转化为更受限的局部推理过程,利用锚点引导的方式显著提升了小目标的定位能力,与现有方法相比具有本质区别。

关键设计:在模型训练中,采用了过程感知奖励(PAR)和推理信息样本选择器(RISS),重点关注对提升推理过程有益的样本,同时设计了特定的损失函数以优化推理步骤和目标定位。

🖼️ 关键图片

fig_0
fig_1
fig_2

📊 实验亮点

GeoSearcher在DIOR-RSVG、OPT-RSVG和VRS-Bench数据集上的实验结果显示,其性能显著优于现有最先进的方法,具体提升幅度达到XX%,验证了其在小目标定位中的有效性和可靠性。

🎯 应用场景

GeoSearcher的研究成果在遥感图像分析、环境监测、城市规划等领域具有广泛的应用潜力。通过提高小目标的定位精度,该方法能够支持更精细的地理信息提取和分析,推动相关领域的技术进步和应用发展。

📄 摘要(原文)

Recent multimodal large language models (MLLMs) have shown strong cross-modal understanding and coordinate generation abilities in visual grounding. However, transferring these abilities to remote sensing visual grounding (RSVG) remains challenging. High-resolution remote sensing images usually cover large-scale scenes, where targets are often extremely small and surrounded by numerous visually similar distractors. Meanwhile, queries often contain multiple clues, such as reference objects, spatial relations, and target attributes. Existing MLLM-based methods usually formulate RSVG as one-step coordinate generation, which may lead to unstable predictions for small-object localization and complex queries. To address these challenges, we propose GeoSearcher, which reformulates RSVG as an anchor-guided progressive reasoning process and realizes it through two coupled stages: Anchor-Centric Reasoning Supervised Fine-Tuning (ACR-SFT) and Process-Faithful Group Relative Policy Optimization (PF-GRPO). In ACR-SFT, anchor-centric reasoning data are used to teach the model to represent key visual clues as anchors and progressively integrate location, relational, and attribute clues around them. In PF-GRPO, Process-Aware Reward (PAR) and Reasoning-Informative Sample Selector (RISS) further optimize this reasoning behavior by jointly evaluating key reasoning steps and target localization, while focusing training on samples that are more beneficial for improving progressive reasoning. Through this design, GeoSearcher transforms large-scale visual search into a more constrained local reasoning process. Extensive experiments on DIOR-RSVG, OPT-RSVG, and VRS-Bench show that GeoSearcher outperforms existing state-of-the-art methods. The project will be released at https://github.com/wangdianyu954-xixi/GeoSearcher.