Hu-Li-Comm/Terahertz-VSAR-Optical-Registration

Video SAR - Optical Image Registration Dataset

1. Dataset Overview

This dataset provides terahertz (THz) video SAR and optical image pairs with certain registration accuracy for cross-modal remote sensing research. It covers multiple typical scenes and SAR imaging modes, and can serve as benchmark data for heterogeneous image registration and matching algorithm validation.

Important Limitation Note: The terahertz video Synthetic Aperture Radar observation data contained in this benchmark dataset are all acquired from airborne field imaging experiments. At present, the research on THz SAR-to-optical image registration is still in an immature stage. Combined with the restrictions brought by the unique physical imaging mechanism of the terahertz band, the original imaging quality and cross-modal registration accuracy of the dataset have objective inherent defects. Researchers can carry out targeted data modification, sample expansion and customized preprocessing based on this dataset to adapt to their differentiated research objectives and experimental schemes.

2. Dataset Composition

2.1 Imaging modes

Spotlight mode and Stripmap mode

2.2 Typical scenes

1. Airport (Spotlight)
2. Sundial (Spotlight)
3. Village (Stripmap)

2.3 Data characteristics

Data form: Continuous video frame sequences with registration correspondence. Each scene contains hundreds of sequential frames, with thousands of valid registration pairs in total, supporting temporal stability analysis of registration algorithms.

3. Application Scenarios

This dataset is applicable to research including:

• Cross-modal image registration and matching
• Multi-source remote sensing data fusion
• Video SAR target tracking
• Heterogeneous image feature extraction and alignment algorithm validation

4. Dataset Structure & File Specification

4.1 Dataset Folder Structure

The dataset is hierarchically organized by SAR acquisition mode, scene type, and cross-modal image modality. Each sample group contains a SAR patch, the matched optical patch, and an optional side-by-side comparison GIF animation.

• Dataset/
  • Spotlight_mode/
    • Airport/
      • Airport_paired_patches/
      • Spotlight_mode_Airport_SAR/
      • Spotlight_mode_Airport_Optical/
      • Spotlight_mode_Airport_Compare/
      • Spotlight_mode_Airport_Compare_Animation.gif
    • Sundial/
      • Sundial_paired_patches/
      • Spotlight_mode_Sundial_SAR/
      • Spotlight_mode_Sundial_Optical/
      • Spotlight_mode_Sundial_Compare/
      • Spotlight_mode_Sundial_Compare_Animation.gif
  • Stripmap_mode/
    • Village/
      • Village_paired_patches/
      • Stripmap_mode_Village_SAR/
      • Stripmap_mode_Village_Optical/
      • Stripmap_mode_Village_Compare/
      • Stripmap_mode_Village_Compare_Animation.gif

4.2 Contents Summary

Mode	Scene	Paired Samples	SAR / Optical Patch Size	Compare Image Size	Index Range
Spotlight mode	Airport	538	2048 × 2048	4096 × 2048	1–538
Spotlight mode	Sundial	312	2048 × 2048	4096 × 2048	001–312
Stripmap mode	Village	2,347	512 × 512	1024 × 512	00001–02347

The full dataset contains 3,197 paired samples, stored as 9,591 PNG files plus 3 GIF animation files.

4.3 Uploaded Sample Previews

We provide partial visual sample files in the repository root, while the full dataset is packaged in the compressed archive.

• Repository Root Files
  • Spotlight mode__Airport
    • Spotlight mode__Airport__SAR__1.png
    • Spotlight mode__Airport__Optical__1.png
    • Spotlight mode__Airport__Compare__1.png
    • Spotlight mode__Airport__Compare__Animation.gif
  • Spotlight mode__Sundial
    • Spotlight mode__Sundial__SAR__001.png
    • Spotlight mode__Sundial__Optical__001.png
    • Spotlight mode__Sundial__Compare__001.png
    • Spotlight mode__Sundial__Compare__Animation.gif
  • Stripmap mode__Village
    • Stripmap mode__Village__SAR__00001.png
    • Stripmap mode__Village__Optical__00001.png
    • Stripmap mode__Village__Compare__00001.png
    • Stripmap mode__Village__Compare__Animation.gif
• Full Dataset Package
  • Terahertz Video SAR-Optical Cross-Modal Image Registration Dataset.zip

Citations

@misc{HuLi_THzVSAR_Optical_Reg_Dataset,

title = {THz Video SAR - Optical Image Registration Dataset},

author = {Hu, Li},

year = {2026},

howpublished = {\url{https://huggingface.co/datasets/Hu-Li-Comm/Terahertz-VSAR-Optical-Registration}},

note = {Apache-2.0}

}