论文数据集

数据主编：向俞明（同济大学） 数据集简介：OSDataset2.0，一个面向SAR-光学影像匹配任务的大规模基准数据集。该数据集包含两部分：局部训练数据集与全幅场景测试集。其中局部训练数据集提供了覆盖全球近20个国家，共6343对512×512像素的配准图像块对（空间分辨率为1米）。数据集由多方法匹配结合人工校验生成。全幅场景测试集提供了一对GeoTIFF格式的整景光学与SAR影像，空间分辨率为0.43米，数据大小为35,783x35,783和35,507x27,298。数据真值基于路灯杆状结构在光学和SAR载荷中的成像机理一致性原则进行采集，共72对高精度真值同名点对。我们同时提供了配套通用评估代码，支持对任意匹配算法进行配准精度的量化分析。 详细使用说明请参考：OSDataset2.0 SAR-光学遥感影像匹配数据集使用说明.pdf 本数据集引用格式 [1]向俞明, 陈锦杨, 洪中华, 等. OSDataset2.0：SAR-光学影像匹配数据集及评估基准[J]. 雷达学报(中英文), 待出版. doi: 10.12000/JR25176 XIANG Yuming, CHEN Jinyang, HONG Zhonghua, et al. OSDataset2.0: SAR-optical image matching dataset and evaluation benchmark[J]. Journal of Radars, 2025, in press. doi: 10.12000/JR25176 Data release time：2025-10-31

Data Editor: Xiang Yuming (Tongji University). Dataset Introduction: OSDataset2.0 is a large-scale benchmark dataset for SAR-optical image matching tasks. The dataset comprises two components: a local training dataset and a full-scene test set. The local training dataset provides 6343 pairs of 512×512 pixel registered image patches with a spatial resolution of 1 m, covering nearly 20 countries worldwide. It was generated via multi-method matching combined with manual verification. The full-scene test set offers a pair of full-scene optical and SAR images in GeoTIFF format, with a spatial resolution of 0.43 m, and the image dimensions are 35,783×35,783 and 35,507×27,298 respectively. The ground truth data were collected based on the consistency principle of the imaging mechanism of lamp-post-like structures in optical and SAR payloads, totaling 72 pairs of high-precision ground-truth corresponding point pairs. We also provide supporting general evaluation code that enables quantitative analysis of registration accuracy for any matching algorithm. For detailed usage instructions, please refer to the document "OSDataset2.0 SAR-optical remote sensing image matching dataset usage instructions.pdf". Citation format of this dataset: [1] XIANG Yuming, CHEN Jinyang, HONG Zhonghua, et al. OSDataset2.0: SAR-optical image matching dataset and evaluation benchmark[J]. Journal of Radars, 2025, in press. doi: 10.12000/JR25176. Data release time: 2025-10-31