Guo et al. (2025) An improved ROBust OpTimization-based (iROBOT) fusion model for reliable spatiotemporal seamless remote sensing data reconstruction

Identification

• Journal: International Journal of Applied Earth Observation and Geoinformation
• Year: 2025
• Date: 2025-11-20
• Authors: Dizhou Guo, Zhenhong Li, Qianqian Jia, Ming Hao
• DOI: 10.1016/j.jag.2025.104964

Research Groups

• State Key Laboratory of Loess Science, Chang’an University, Xi’an, China
• College of Geological Engineering and Geomatics, Chang’an University, Xi’an, China
• Key Laboratory of Western China’s Mineral Resources and Geological Engineering, Ministry of Education, Xi’an, China
• Key Laboratory of Ecological Geology and Disaster Prevention, Ministry of Natural Resources, Xi’an, China
• School of Geosciences and Info-physics, Central South University, Changsha, China
• Jiangsu Key Laboratory of Resources and Environmental Information Engineering, China University of Mining and Technology, Xuzhou, China

Short Summary

This study introduces iROBOT, an improved spatiotemporal fusion model that addresses block artifacts and cloud contamination in remote sensing data reconstruction by employing object-level processing and an adaptive gap-filling strategy. Experiments demonstrate iROBOT's superior accuracy and robustness compared to existing methods, particularly in cloud-prone environments.

Objective

• To overcome the limitations of the ROBOT fusion model, specifically the generation of block artifacts due to its sliding-window strategy and degraded reconstruction quality under cloud-contaminated conditions, by proposing an improved version (iROBOT).

Study Configuration

• Spatial Scale: Landsat 8 surface reflectance imagery (fine resolution, typically 30 meters) and MODIS MCD43A4 imagery (coarse resolution, typically 500 meters, resampled to Landsat's spatial extent). Study areas covered 800 pixels × 800 pixels.
• Temporal Scale: Time-series remote sensing imagery from 2019 to 2022, with specific prediction phases and auxiliary images spanning days to months.

Methodology and Data

• Models used: iROBOT (proposed), ROBOT (Robust Optimization-based), FSDAF (Flexible Spatiotemporal Data Fusion), OL-HSTFM (Object-Level Hybrid Spatiotemporal Fusion Method).
• Data sources: Landsat 8 surface reflectance imagery (fine imagery), MODIS MCD43A4 imagery (coarse imagery).

Main Results

• iROBOT consistently outperformed ROBOT, OL-HSTFM, and FSDAF in both cloud-free and cloud-contaminated scenarios across two representative study areas (Poyang Lake Wetland Area and Tianshan North Foothills Agricultural Area).
• Compared to ROBOT, iROBOT reduced RMSE by 20.3% and 30.4%, increased SSIM by 5.8% and 4.0%, and increased PSNR by 8.3% and 8.7% in the Poyang Lake Wetland Area and Tianshan North Foothills Agricultural Area, respectively.
• The proposed object-level processing effectively suppressed block artifacts and better preserved structural details compared to ROBOT's moving-window strategy.
• The adaptive gap-filling strategy and low-quality information detection module significantly enhanced fusion reliability under cloud contamination.
• iROBOT demonstrated superior capability in utilizing multi-temporal auxiliary information, with its accuracy advantage becoming more pronounced as the number of auxiliary inputs increased, especially in cloud-prone environments.
• iROBOT maintained high computational efficiency, being 20.8–21.5 times faster than FSDAF and 2.0–2.7 times faster than OL-HSTFM.

Contributions

• Introduction of iROBOT, an improved spatiotemporal fusion model that significantly enhances the reliability and accuracy of remote sensing data reconstruction.
• Development of an object-level processing strategy that replaces fixed rectangular patches with spectrally and spatially homogeneous segments, effectively eliminating block artifacts and preserving structural details.
• Implementation of an adaptive gap-filling strategy for cloud-contaminated temporal nearest-neighbor auxiliary images and prediction-phase coarse images.
• Integration of a low-quality information detection module to filter unreliable cloud-contaminated data from regression and residual allocation steps, thereby improving robustness.
• Demonstrated superior performance over state-of-the-art methods (ROBOT, FSDAF, OL-HSTFM) in diverse conditions, including cloud-contaminated and cloud-free scenarios, and across different land cover types.
• Highlighted iROBOT's potential for reliable seamless global data cube reconstruction due to its enhanced accuracy, robustness, and computational efficiency.

Funding

• National Natural Science Foundation of China (42501436, 42271368)
• Postdoctoral Fellowship Program of CPSF (GZC20250163)
• Natural Science Basic Research Plan in Shaanxi Province of China (2025JC-YBQN-408)
• Shaanxi Province Postdoctoral Science Foundation (2024BSHSDZZ227)
• Fundamental Research Funds for the Central Universities, CHD (300102265105)
• Shaanxi Province Geoscience Big Data and Geohazard Prevention Innovation Team (2022)
• Generic Technical Development Platform of Shaanxi Province for Imaging Geodesy (2024ZG-GXPT-07)

Citation

@article{Guo2025improved,
  author = {Guo, Dizhou and Li, Zhenhong and Jia, Qianqian and Hao, Ming},
  title = {An improved ROBust OpTimization-based (iROBOT) fusion model for reliable spatiotemporal seamless remote sensing data reconstruction},
  journal = {International Journal of Applied Earth Observation and Geoinformation},
  year = {2025},
  doi = {10.1016/j.jag.2025.104964},
  url = {https://doi.org/10.1016/j.jag.2025.104964}
}