Hybrid Optical and Synthetic Aperture Radar (SAR) Ship Re-Identification Dataset (HOSS ReID dataset)
收藏资源简介:
HOSS ReID数据集是一个混合光学和合成孔径雷达(SAR)的船只再识别数据集,旨在评估使用低地球轨道的混合光学和SAR传感器星座进行船只跟踪的有效性。该数据集由13个图像序列组成,包含18张SAR图像和25张光学图像,总计43帧。数据集中的图像已经过几何校正和辐射校正,但未使用数字高程模型(DEM)进行正射校正。光学图像的地面采样距离(GSD)为0.75米,而SAR图像的GSD为1米。数据集分为训练集和验证集,并包含163个干扰对象,以增强数据集的复杂性和模拟现实世界场景。
The HOSS ReID Dataset is a vessel re-identification dataset combining optical and Synthetic Aperture Radar (SAR) data, which aims to evaluate the effectiveness of vessel tracking using a constellation of mixed optical and SAR sensors in Low Earth Orbit (LEO). This dataset comprises 13 image sequences, containing 18 SAR images and 25 optical images, with a total of 43 frames. All images in the dataset have undergone geometric and radiometric corrections, but no orthorectification using a Digital Elevation Model (DEM) has been applied. The Ground Sampling Distance (GSD) of optical images is 0.75 meters, while that of SAR images is 1 meter. The dataset is split into training and validation sets, and includes 163 distractor objects to enhance the complexity of the dataset and simulate real-world scenarios.
HOSS ReID数据集概述
数据集简介
- 名称:HOSS ReID数据集(Hybrid Optical and Synthetic Aperture Radar Ship Re-Identification Dataset)
- 用途:评估利用低地球轨道光学和SAR传感器星座进行船舶跟踪的有效性
- 特点:包含同一船舶在不同条件下长时间拍摄的图像,使用不同模态的卫星在不同时间和角度拍摄
数据集内容
-
数据模态:光学和合成孔径雷达(SAR)图像
-
数据获取方式:通过不同卫星在不同时间、角度和条件下拍摄同一船舶的图像
-
数据存储结构:
data └── HOSS └── bounding_box_test ..
数据集优势
- 缩短重成像周期
- 实现全天候跟踪
- 适用于实际船舶跟踪需求
数据集获取
- 公开获取地址:https://zenodo.org/records/15751352
相关研究
- 配套方法:TransOSS(基于Vision Transformer架构的跨模态船舶重识别方法)
- 论文地址:https://arxiv.org/abs/2506.22027
引用格式
bibtex @misc{wang2025crossmodal, title={Cross-modal Ship Re-Identification via Optical and SAR Imagery: A Novel Dataset and Method}, author={Han Wang and Shengyang Li and Jian Yang and Yuxuan Liu and Yixuan Lv and Zhuang Zhou}, year={2025}, eprint={2506.22027}, archivePrefix={arXiv}, primaryClass={cs.CV}, url={https://arxiv.org/abs/2506.22027}, }




