Image Matching via Pre-deformed Reference Images
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Traditional feature detection algorithms fail to robustly match images captured at large tilt angles due to distortion, leading to deformation and a disproportionate aspect ratio, resulting in a high failure rate of image matching.
Innovation Solution
The method extends the reference image by deforming it into multiple extended reference images through compression or magnification in various directions, allowing for improved matching with distorted detected images, and uses a homography matrix to select the optimal reference image based on the imaging direction angle, reducing calculation complexity and enabling real-time image matching.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional feature detection algorithms are used for image matching, then the algorithm is simple and fast, but the matching fails when images are captured at large tilt angles due to distortion
Solution Approach 1:
The patent applies preliminary action by pre-generating multiple candidate reference images with different distortion levels and types before the actual matching process. When a detected image is captured, the system selects and compares against the pre-prepared candidate images that match the distortion characteristics, avoiding the need for complex real-time distortion correction and enabling reliable matching at large tilt angles
Solution Approach 2:
The patent changes the parameter of reference images by creating multiple versions with varying distortion parameters (different tilt angles, aspect ratios, and deformation levels). This allows the system to adapt to different captured image conditions by selecting the appropriate reference image variant, thereby improving robustness to image distortion without complicating the matching algorithm itself
2Reliability
If multiple extended reference images are generated to handle distortion, then the matching robustness improves, but the calculation complexity increases
Solution Approach 1:
The patent reduces calculation complexity by performing the computationally intensive image generation and distortion application in advance, before runtime. The candidate reference images are pre-processed and stored, so during actual matching, the system only needs to compare the detected image against these pre-generated candidates, significantly reducing real-time computational load while maintaining high matching success rates
Solution Approach 2:
The patent applies partial action by generating only the specific number and types of candidate reference images that are most likely to be needed based on expected operating conditions. Rather than exhaustively generating all possible distortion variations, the system creates a focused set of candidate images covering the most probable tilt angles and distortion levels, balancing robustness with computational efficiency
Data Source
Figure 1A~1C
Figure 2~3
Figure 4
AI summary
This application provides an image matching method, an image matching apparatus, an image matching system, and a storage medium, the image matching method comprises: an extension step (S201) of extending an original reference image into one or more extended reference images, wherein the one or more extended reference images is obtained by deforming the original reference image; and an image matching step (S202) of image-matching a detected image with candidate reference images that include the original reference image and the one or more extended reference images.