IRSatVideo Dataset
收藏资源简介:
IRSatVideo是一个用于低地球轨道和地球静止轨道卫星视频中红外小目标检测的大规模半模拟数据集,包含真实卫星背景图像和合成的卫星运动、目标外观、轨迹和强度。IRSatVideo-LEO基于Landsat 7-8开发,包含200个图像序列和91021帧;IRSatVideo-GEO基于高分4号开发,包含200个图像序列和94136帧,提供实例级分割标注作为红外卫星视频基准
IRSatVideo is a large-scale semi-simulated dataset for infrared small target detection in satellite videos from low Earth orbit (LEO) and geostationary orbit (GEO). It includes real satellite background images, as well as synthetic satellite motions, target appearances, trajectories and target intensities. The IRSatVideo-LEO subset, developed based on Landsat 7-8, contains 200 image sequences and 91,021 frames; the IRSatVideo-GEO subset, developed based on Gaofen-4 (GF-4), includes 200 image sequences and 94,136 frames. This dataset provides instance-level segmentation annotations, serving as a benchmark for infrared satellite video target detection.
IRSatVideo数据集概述
数据集基本信息
- 数据集名称:IRSatVideo Dataset
- 数据集类型:半模拟数据集
- 应用领域:红外小目标检测
- 应用场景:LEO和GEO卫星视频中的多帧红外小目标检测
数据集构成
IRSatVideo-LEO
- 数据来源:Landsat 7-8卫星
- 数据规模:200个图像序列,共91021帧
IRSatVideo-GEO
- 数据来源:GaoFen 4卫星
- 数据规模:200个图像序列,共94136帧
数据特征
- 背景图像:真实卫星背景图像
- 模拟内容:合成的卫星运动、目标外观、轨迹和强度
- 标注类型:实例级分割标注
- 用途:为MIRST检测和跟踪提供红外卫星视频基准
数据格式
目录结构
- images:图像数据目录
- masks:掩码数据目录
- video_idx:视频索引文件
- img_idx:图像索引文件
- voc:VOC格式标注文件
索引文件
IRSatVideo-LEO
- train_IRSatVideo-LEO.txt
- test_IRSatVideo-LEO.txt
- test_IRSatVideo-LEO-easy.txt
- test_IRSatVideo-LEO-middle.txt
- test_IRSatVideo-LEO-hard.txt
IRSatVideo-GEO
- train_IRSatVideo-GEO.txt
- test_IRSatVideo-GEO.txt
- test_IRSatVideo-GEO-SCR1.txt
- test_IRSatVideo-GEO-SCR2.txt
- test_IRSatVideo-GEO-SCR3.txt
获取方式
- 访问表单:https://forms.cloud.microsoft/r/eYC3mNz5s3
相关论文
- 论文链接:https://arxiv.org/abs/2409.12448
联系方式
- 联系邮箱:yingxinyi18@nudt.edu.cn
引用格式
bibtex @article{RFR, author = {Xinyi Ying, Li Liu, Zaipin Lin, Yangsi Shi, Yingqian Wang, Ruojing Li, Xu Cao, Boyang Li, Shilin Zhou}, title = {Infrared Small Target Detection in Satellite Videos: A New Dataset and A Novel Recurrent Feature Refinement Framework}, journal = {IEEE Transactions on Geoscience and Remote Sensing}, year = {2025}, }




