Remote Sensing Image Matching Using Oriented Self-Similar Features
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing multi-modal remote sensing image matching techniques face challenges in balancing geometric invariance with high-precision registration due to signal-to-noise ratio differences and geometric transformation issues, leading to inaccurate feature matching and high computational complexity.
Innovation Solution
A multi-modal remote sensing image hybrid matching method utilizing multi-dimensional oriented self-similar features, comprising blended feature coarse matching and multi-dimensional oriented self-similar feature fine matching, which includes steps for feature extraction, geometric transformation, and enhanced template feature matching using three-dimensional phase correlation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If template matching methods are used to achieve high accuracy in identical point recognition, then matching precision is improved, but computational complexity increases significantly
Solution Approach 1:
The patent divides the matching process into two distinct stages: coarse matching using feature-based methods to identify candidate regions, and fine matching using template matching only in those restricted regions. This segmentation allows template matching to be applied selectively rather than globally, maintaining high precision where needed while reducing overall computational complexity.
Solution Approach 2:
The patent applies template matching partially rather than exhaustively across the entire image. By limiting template matching operations to small candidate regions identified by the coarse feature matching stage, the method achieves sufficient precision for the application while avoiding the excessive computational cost of full-image template matching.
2Adaptability or versatility
If feature-based methods are used to address geometric transformation differences, then adaptability to geometric transformations is improved, but matching precision deteriorates due to signal-to-noise ratio differences
Solution Approach 1:
The patent segments the matching task into two phases with different methods optimized for different requirements. The first phase uses feature-based methods that are robust to geometric transformations, while the second phase uses template matching that provides higher precision for the final alignment, thus combining the strengths of both approaches.
Solution Approach 2:
The patent uses feature-based matching results as an intermediary step to generate candidate regions and initial transformation parameters. These intermediates guide the subsequent template matching process, allowing the system to benefit from both the geometric robustness of feature methods and the precision of template methods.
3Measurement precision
If pixel-by-pixel matching strategy is used to achieve accurate location of identical points, then matching precision is improved, but computational complexity increases
Solution Approach 1:
The patent divides the image space into candidate regions identified by feature matching and non-candidate regions. Template matching is applied only within the segmented candidate regions, achieving accurate location of identical points where needed while avoiding unnecessary computations in other areas, thus improving computational efficiency.
Solution Approach 2:
The patent applies the computationally intensive pixel-by-pixel template matching strategy partially only in candidate regions rather than across the entire image. This partial application achieves the necessary location accuracy for matching while significantly reducing the total computational burden compared to exhaustive pixel-by-pixel matching.
Data Source
AI summary
The present invention provides a method and system for multi-modal remote sensing image hybrid matching with multi-dimensional oriented self-similarity features. The method comprises: hybrid feature coarse matching is executed by rapidly extracting the image's self-similarity features using the offset mean filtering method and describing the feature points using the directional information of the self-similarity features. Through this coarse matching, an affine transformation model between multi-modal images is estimated, and an initial affine transformation is applied to the matched image. Next, the multi-dimensional oriented self-similarity feature fine matching is performed by constructing multi-dimensional oriented self-similarity template features using the multi-channel self-similarity map obtained from the hybrid feature coarse matching stage. Through a subsampling strategy applied to the constructed template features and convolution enhancement using a three-dimensional Gaussian kernel, the directional self-similarity features are strengthened. Finally, a three-dimensional phase consistency measure is used to identify highly accurate matching homologous points.