This is a set of groundtruth on satellite video for evaluating moving object detection algorithm.
In this project, we annotate the moving vehicles on two satellite videos, each of whom contains 700 frames. The geo-territories of them are shown as below (Red for 001, and Green for 002).
For each moving vehicle, a boundary box is provided across the video with an unique id, therefore this dataset can also be used for evaluating multiple target tracking algorithm.
The boundary boxes are annotated on Computer Vision Annotation Tool (CVAT).
Please refer to the example file for more details.
If you use this library for your publications, please cite it as:
@ARTICLE{8930094,
author={J. {Zhang} and X. {Jia} and J. {Hu}},
journal={IEEE Transactions on Geoscience and Remote Sensing},
title={Error Bounded Foreground and Background Modeling for Moving Object Detection in Satellite Videos},
year={2019},
volume={},
number={},
pages={1-11},
doi={10.1109/TGRS.2019.2953181},
ISSN={1558-0644},
month={},}
Additional refereence:
@ARTICLE{9037205,
author={J. {Zhang} and X. {Jia} and J. {Hu} and J. {Chanussot}},
journal={IEEE Transactions on Geoscience and Remote Sensing},
title={Online Structured Sparsity-Based Moving-Object Detection From Satellite Videos},
year={2020},
volume={},
number={},
pages={1-14},
doi={10.1109/TGRS.2020.2976855},
ISSN={1558-0644},
month={},}
@INPROCEEDINGS{8615873,
author={J. {Zhang} and X. {Jia} and J. {Hu} and K. {Tan}},
booktitle={2018 Digital Image Computing: Techniques and Applications (DICTA)},
title={Satellite Multi-Vehicle Tracking under Inconsistent Detection Conditions by Bilevel K-Shortest Paths Optimization},
year={2018},
volume={},
number={},
pages={1-8},
doi={10.1109/DICTA.2018.8615873},
ISSN={null},
month={Dec},}
Please refer to LICENSE for more details.