A research team from Wuhan University has developed a new artificial intelligence framework that can restore partially hidden objects in satellite imagery more reliably than existing methods. The framework, called Remote Sensing Amodal Completion (RSAC), infers complete object shape, surface texture, and semantic identity from incomplete observations, rather than merely filling missing pixels. This approach addresses a key challenge in geospatial AI: how to reconstruct complete ground objects when only fragments are visible due to cloud cover, overlapping objects, or imaging angles.
The study, published in the Journal of Remote Sensing on April 7, 2026, proposes a Dual-Adaptive Diffusion-Based Framework that adapts Stable Diffusion to the remote sensing domain using Low-Rank Adaptation (LoRA). A four-channel ControlNet uses image and mask information to guide structural completion, while a prior-enhanced initialization strategy preserves low-frequency information from visible objects. Compared to other methods like Stable Diffusion Inpainting and LaMa, the proposed framework produced more accurate object geometry, clearer boundaries, and more realistic texture continuity.
Satellite imagery is critical for disaster response, urban planning, environmental monitoring, and security analysis, but objects are frequently obscured. Existing inpainting methods often produce visually plausible results that distort object structure or hallucinate incorrect content. The RSAC framework shifts from scene-level inpainting to object-level reasoning, helping machines infer what an object is and how it should be structured. The team built a dedicated dataset of 1,770 annotated instances across 10 categories, including planes, ships, and sports fields. In tests, the method achieved 100% valid-output coverage, an Intersection over Union of 0.853, and a structural similarity index of 0.930, outperforming baseline methods.
The technology could support more reliable geospatial intelligence in scenarios like post-disaster assessment, infrastructure mapping, and urban monitoring. By restoring complete object morphology from partial observations, RSAC may also improve training data for detection models and help AI systems interpret satellite imagery more like human analysts. The research was supported by the National Natural Science Foundation of China under grants 42422109 and 42371366. Future studies may extend the framework to more object categories, dynamic drone perspectives, and 3D reconstruction.
