A new artificial intelligence framework developed by researchers at Wuhan University offers a more reliable method for reconstructing partially hidden objects in satellite imagery. Rather than simply filling in missing pixels, the framework infers complete object shape, surface texture, and semantic identity from incomplete observations, addressing a critical challenge in geospatial artificial intelligence.
Satellite imagery is essential for disaster response, urban planning, environmental monitoring, and security analysis. However, ground objects are frequently obscured by clouds, overlapping structures, or imaging angles, leading to misclassification and fragmented mapping. Existing inpainting methods often produce visually plausible results but can distort object structure or generate incorrect content. The new framework, detailed in the Journal of Remote Sensing (DOI: 10.34133/remotesensing.1035), introduces Remote Sensing Amodal Completion (RSAC) as a dedicated task for reconstructing complete objects from partial observations.
The proposed Dual-Adaptive Diffusion-Based Framework adapts Stable Diffusion through Low-Rank Adaptation (LoRA) and uses a four-channel ControlNet to guide structural completion. A prior-enhanced initialization strategy preserves low-frequency information from visible portions, improving physical consistency. In comparative tests against methods like Stable Diffusion Inpainting and LaMa, the framework achieved superior geometry, clearer boundaries, and realistic texture continuity. It achieved 100% valid-output coverage, an Intersection over Union of 0.853, and a structural similarity index of 0.930.
The researchers built a dedicated dataset with 1,770 annotated instances across 10 categories, including planes, ships, and sports fields. The framework also enhanced downstream object detection and supported layered 2.5D scene understanding. The team emphasized that the goal is to help machines infer object identity and structure, not just visual completeness. This technology could improve geospatial intelligence in scenarios like post-disaster assessment, infrastructure mapping, and urban monitoring. Future work may extend to more object categories, drone perspectives, and multimodal data.
