A planet remote sensing image feature matching method, system, device and medium combining spatial and frequency domain edge structure information
By combining spatial and frequency domain edge structure information, a weighted phase consistency maximum moment map and an adaptive gradient map are constructed to optimize planetary remote sensing image feature matching. This solves the matching difficulties caused by nonlinear radiometric differences and achieves high-precision and stable image registration.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- TONGJI UNIV
- Filing Date
- 2026-03-18
- Publication Date
- 2026-07-17
AI Technical Summary
Existing planetary remote sensing image feature matching methods are unable to address the significant nonlinear radiation differences caused by factors such as solar elevation angle, observation angle, and sensor response characteristics. This leads to local brightness reversal, shadow changes, and loss of texture details in the images. Repetitive texture structures interfere with feature matching, reducing matching accuracy and precision.
By combining spatial and frequency domain edge structure information, a weighted phase consistency maximum moment map is constructed to perform scale-invariant feature detection, generating a spatial adaptive multi-scale first-order gradient map and a frequency domain weighted phase consistency maximum moment map. The nearest neighbor matching strategy is combined with fast sample consistency mismatch elimination, and a global-to-local guided matching strategy is introduced to optimize the matching results.
It improves the accuracy and robustness of feature matching for planetary remote sensing images, adapts to complex terrain and light variations, reduces mismatches, and enhances matching accuracy and reliability, especially in image registration in complex environments such as the Moon and Mars.
Smart Images
Figure CN122416072A_ABST