发表机构
Eurecat; IIE, Facultad de Ingeniería, Universidad de la República; Universitat Pompeu Fabra; Université Paris-Saclay; CNRS; ENS Paris-Saclay; Institut Universitaire de France（欧罗卡特技术中心; 乌拉圭共和国大学工程学院IIE; 庞培法布拉大学; 巴黎萨克雷大学; 法国国家科学研究中心; 巴黎萨克雷高等师范学校; 法国大学研究院）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
SeasonStereo框架利用带可控季节外观变化的合成图像对训练并结合基础模型零样本几何先验，实现低成本、高精度的多日期卫星图像密集立体匹配，支持大规模三维重建。
AI 中文摘要
从卫星图像进行精确三维重建通常依赖近同步立体像对，这限制了其在多日期图像存在季节和光照条件变化的历时场景中的适用性。训练对外观变化鲁棒的密集立体匹配模型是长期存在的挑战，因为大规模获取对齐的多日期图像和真实几何信息成本高昂。我们提出SeasonStereo，这是一个可扩展框架，通过在具有可控季节外观变化的合成图像对上训练，同时利用基础模型的零样本几何先验，解决多日期卫星图像的视差估计问题。SeasonStereo达到了最先进的LiDAR监督模型的精度，同时产生更清晰的几何细节，无需对齐的真实多日期训练产品或LiDAR衍生标签。因此，SeasonStereo为从异构卫星图像进行大规模三维重建提供了实用路径，降低了监督成本。
英文摘要
Accurate 3D reconstruction from satellite imagery typically relies on near-simultaneous stereo pairs, limiting its applicability to diachronic settings where multi-date images exhibit varying seasonal and illumination conditions. Training dense stereo matching models robust to appearance changes is a long-standing challenge, as aligned multi-date imagery and ground-truth geometry are costly to obtain at scale. We propose SeasonStereo, a scalable framework that addresses disparity estimation from diachronic satellite images by training on synthetic image pairs with controlled seasonal appearance variation, while leveraging zero-shot geometric priors from foundation models. SeasonStereo matches the accuracy of state-of-the-art LiDAR-supervised models, while producing sharper geometric details without requiring aligned real multi-date training products or LiDAR-derived labels. As a result, SeasonStereo offers a practical path toward large-scale 3D reconstruction from heterogeneous satellite images with reduced supervision cost.
