Collect 10-year time-series remote sensing data of typical mobile sand dune areas by searching historical images of World Imagery Wayback. Based on typical surface features such as clear sand ridges and obvious texture changes in high-resolution images, the U-Net semantic segmentation model of deep learning was used to extract the position of sand ridges in each period. Combined with optical flow method, EWMA method, and visual interpretation, the rapid change area of sand ridges was comprehensively determined to verify the spatial authenticity and accuracy of sand dune morphology changes in temporal images. Finally, a time-series sand ridge line and velocity variation area dataset covering typical mobile sand dunes were prepared, providing reliable data support for the study of sand dune movement processes and velocity variation analysis.
World Imagery Wayback imagery: accessed through ArcGIS Wayback Imagery service
1. Semantic segmentation and extraction of sand ridges: Utilizing the deep learning U-Net semantic segmentation model to automatically extract sand ridges at different stages. Based on the typical geomorphic features presented by the sand ridge line in the image, such as obvious texture changes, clear brightness differences, and continuous overall direction, the quality of the model output is verified to obtain accurate temporal sand ridge line results that depict the geometric shape of sand dunes, forming the core label part of the dataset.
2. Optical flow method and EWMA detection of rapidly changing areas and visual verification: Based on the extraction of sand ridge lines, optical flow method combined with EWMA control map is used to detect the rapidly changing areas of sand dune movement, and visual interpretation is carried out through high-resolution images to verify the detection results, ensuring that the detected rapidly changing areas are consistent with the actual rapidly changing areas of sand dunes.
Data quality verification shows that the distribution of sand ridge lines is consistent with the original image morphology, and the labeling of rapidly changing areas is accurate and reliable, providing a solid foundation for sand dune movement analysis.
| # | number | name | type |
| 1 | 2022YFF0711700 | National key R & D plan | |
| 2 | 2022YFF0711703 | Permafrost Variable Rapid Change Intelligent Discovery and Evolution Perception Data Engineering | National key R & D plan |
This work is licensed under
CC BY 4.0 (Creative Commons Attribution 4.0 International License).
CSTR:11738.11.NCDC.NIEER.DB7027.2025
10.12072/ncdc.nieer.db7027.2025
