An image feature matching method based on high-order graph global consistency
By constructing a high-order graph structure and using an information entropy adaptive filtering method, the feature matching problem of remote sensing images in scenarios with high outliers and complex geometric distortions was solved, achieving image registration with high robustness and high adaptability.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- XIAMEN UNIV OF TECH
- Filing Date
- 2026-05-20
- Publication Date
- 2026-07-17
AI Technical Summary
Existing image feature matching methods perform poorly in scenarios with high outlier rates and complex geometric distortions, making it difficult to accurately match remote sensing images, especially limiting the cross-scene generalization ability of remote sensing images.
The image feature matching method based on high-order graph global consistency constructs an initial graph structure, generates graph structures of different orders and fuses them, uses information entropy to adaptively filter out noisy connections, guides geometric transformation model estimation, generates a high-confidence global representation graph, and selects the correct matching set.
Achieving robust feature matching in scenarios with high outlier counts and complex geometric transformations improves the accuracy and adaptability of remote sensing image registration, reduces reliance on large-scale labeled data, and adapts to different sensors and imaging conditions.
Smart Images
Figure CN122416074A_ABST