Li et al. (2026) Physics-Prior-Guided Feature Pyramid Network for Unified Multi-Angle Spectral–Polarimetric Cloud Detection

⚠️ Warning: This summary was generated from the abstract only, as the full text was not available.

Identification

• Journal: Remote Sensing
• Year: 2026
• Date: 2026-04-12
• Authors: S. M. Li, Xingyuan Ji, Xiaoxue Chu, Song Ye, Ziyang Zhang, Yongyin Gan, Xiangyu Wang, Fangyuan Wang
• DOI: 10.3390/rs18081150

Research Groups

Not explicitly stated in the provided text.

Short Summary

This study proposes a novel deep learning framework, the Multi-angle Polarization Feature Pyramid Structure (MP-FPS), to enhance cloud detection by leveraging joint spectral analysis and multi-angle polarization data. Evaluated on the global POLDER-3 dataset, MP-FPS achieves a mean Intersection over Union (mIoU) of 0.8662, surpassing the official baseline by 12.4%.

Objective

• To develop a robust cloud detection framework that resolves spectral ambiguities between clouds and bright or heterogeneous surfaces by explicitly leveraging joint spectral features and multi-angle polarization data.

Study Configuration

• Spatial Scale: Global (evaluated on the global POLDER-3 dataset).
• Temporal Scale: Not explicitly defined for the study, but evaluated on a dataset derived from a satellite mission (POLDER-3).

Methodology and Data

• Models used: Multi-angle Polarization Feature Pyramid Structure (MP-FPS), a novel deep learning framework featuring a dual-branch network for modality disentanglement and fusion, a hierarchical multi-scale cross-channel multi-angle fusion module, and a channel-space dual-path attention mechanism.
• Data sources: Global POLDER-3 dataset, which provides multi-angle and polarization data.

Main Results

• The proposed MP-FPS framework achieved a mean Intersection over Union (mIoU) of 0.8662 for cloud detection.
• MP-FPS surpassed the official baseline by 12.4% in mIoU.
• The framework significantly improved detection accuracy in challenging regions, including cloud edges and thin cirrus.
• Joint spectral analysis was identified as a critical enabler for high-precision cloud masking.

Contributions

• Demonstrated the underexploited discriminative power of joint spectral analysis for robust cloud detection, particularly in ambiguous surface conditions.
• Proposed MP-FPS, a novel deep learning framework that explicitly integrates joint spectral features with multi-angle polarimetric information.
• Introduced a dual-branch network for adaptive fusion of spectral and multi-angle polarization modalities.
• Developed a hierarchical, multi-scale cross-channel multi-angle fusion module to capture spatial–spectral–angular dependencies.
• Implemented a channel-space dual-path attention mechanism to refine sub-pixel responses, enhancing accuracy in challenging cloud regions.
• Achieved state-of-the-art performance on the global POLDER-3 dataset, setting a new benchmark for cloud detection accuracy.

Funding

Not explicitly stated in the provided text.

Citation

@article{Li2026PhysicsPriorGuided,
  author = {Li, S. M. and Ji, Xingyuan and Chu, Xiaoxue and Ye, Song and Zhang, Ziyang and Gan, Yongyin and Wang, Xiangyu and Wang, Fangyuan},
  title = {Physics-Prior-Guided Feature Pyramid Network for Unified Multi-Angle Spectral–Polarimetric Cloud Detection},
  journal = {Remote Sensing},
  year = {2026},
  doi = {10.3390/rs18081150},
  url = {https://doi.org/10.3390/rs18081150}
}