Heterogeneous SAR-optical fusion for near-real-time land use and land cover mapping under cloud contamination: A novel framework and global benchmark dataset

2026-06-16 · Source: Computer Vision and Pattern Recognition · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Data Science & Analytics · Depth: Expert, quick

Summary

CloudLULC-Net is a novel end-to-end heterogeneous SAR-optical fusion framework for near-real-time land use and land cover (LULC) mapping. It directly predicts LULC maps from cloud-contaminated Sentinel-2 optical imagery and Sentinel-1 SAR observations, specifically addressing cloud contamination issues. The framework incorporates optical reliability modulation, heterogeneous information adaptive aggregation, and a unified semantic mapping transformer. A semantic anchor-guided optimization strategy improves intermediate semantic representation consistency. To support this, the CloudLULC-Set benchmark dataset was created. It contains 40,223 SAR-optical-label triplets with pixel-level LULC annotations across diverse regions and cloud conditions. CloudLULC-Net achieved an OA of 86.60%, an F1-score of 83.29%, and an mIoU of 73.51%. These results demonstrate its robustness and practical value, outperforming representative existing methods for target-date LULC mapping in cloud-prone regions.

Key takeaway

If you are a remote sensing analyst or GIS professional mapping land use and land cover (LULC) in cloud-prone regions, consider CloudLULC-Net. This SAR-optical fusion framework directly addresses optical imagery limitations from cloud contamination. It provides higher accuracy (OA 86.60%) than traditional methods, enabling more consistent target-date LULC mapping. Explore the public project at https://github.com/RSIIPAC/CloudLULC for implementation details and to integrate this robust solution.

Key insights

CloudLULC-Net fuses SAR and cloud-contaminated optical data for robust LULC mapping, overcoming traditional cloud limitations.

Principles

• Modulate optical reliability to suppress noise.
• Aggregate heterogeneous data adaptively.
• Use semantic anchors for representation consistency.

Method

CloudLULC-Net uses optical reliability modulation, heterogeneous information adaptive aggregation, and a unified semantic mapping transformer, optimized with a semantic anchor-guided strategy.

In practice

• Map LULC in cloud-prone regions.
• Utilize Sentinel-1 and Sentinel-2 data.
• Improve accuracy over existing global products.

Topics

• SAR-Optical Fusion
• Land Use Land Cover Mapping
• Cloud Contamination
• Sentinel-1
• Sentinel-2
• CloudLULC-Net
• Deep Learning

Code references

• RSIIPAC/CloudLULC

Best for: Computer Vision Engineer, AI Scientist, Machine Learning Engineer, Research Scientist

Related on AIssential

• deFOREST: Fusing Optical and Radar satellite data for Enhanced Sensing of Tree-loss — Fusing optical and SAR satellite data with a novel anomaly detection method enhances deforestation monitoring, especially in cloudy areas.
• Geometry-Aware Cross Modal Alignment for Light Field-LiDAR Semantic Segmentation — Integrating light field images and LiDAR point clouds significantly enhances semantic segmentation, especially for occluded objects.
• RSRCC: A Remote Sensing Regional Change Comprehension Benchmark Constructed via Retrieval-Augmented Best-of-N Ranking — RSRCC is a new benchmark for fine-grained, localized remote sensing change question-answering.
• Global Offshore Wind Infrastructure: Deployment and Operational Dynamics from Dense Sentinel-1 Time Series — A new global Sentinel-1 SAR dataset enables high-resolution monitoring of offshore wind infrastructure deployment and operations.
• RPC-GS: Gaussian Splatting with native RPC Rendering for Satellite Imagery — Native RPC integration in Gaussian Splatting significantly improves satellite imagery 3D reconstruction accuracy.
• Heterogeneous LiDAR Early Fusion and Learned Re-Ranking Strategy for Robust Long-Term Place Recognition in Unstructured Environments — Heterogeneous LiDAR early fusion and learned re-ranking significantly boost place recognition in challenging unstructured agricultural environments.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Computer Vision and Pattern Recognition.