Fundamental Matrix Generation Using Derived Feature Points
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for generating a fundamental matrix require detecting multiple pairs of feature points between images, which can be cumbersome and inefficient, especially when internal camera parameters like focal length are unknown.
Innovation Solution
A system that detects three or more feature point pairs between images and generates a fundamental matrix using derived point pairs, where the distances and directions for derivation are determined based on computed feature values, allowing for the generation of a fundamental matrix with fewer required feature point pairs.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If multiple pairs of feature points are detected between images using conventional methods, then the fundamental matrix can be generated with sufficient accuracy, but the process becomes cumbersome and inefficient
Solution Approach 1:
The patent segments the feature point utilization process into two stages: first detecting multiple feature point pairs between images, then generating additional derived point pairs from these detected points. This segmentation allows the system to reuse the initially detected feature points to create multiple additional corresponding points through geometric transformations, thereby improving efficiency without sacrificing the accuracy that would require detecting many more original feature points
Solution Approach 2:
The patent performs preliminary detection of feature point pairs between images, then uses these detected points as a basis to generate derived point pairs through geometric transformations (translation, rotation, scaling). This preliminary action of detecting feature points once and reusing them multiple times for deriving additional points eliminates the need to detect many more feature points, thus improving productivity while maintaining measurement precision
2Adaptability or versatility
If feature point pairs are detected from images with unknown camera parameters, then the fundamental matrix can be estimated, but the complexity of the detection process increases
Solution Approach 1:
The patent changes the parameter representation by working with derived point pairs generated through geometric transformations rather than requiring direct detection of all necessary correspondence points. By transforming the problem from detecting many feature points to generating points through parameter-based transformations (translation vectors, rotation angles, scaling factors), the system adapts to unknown camera parameters while simplifying the overall detection process complexity
3Measurement precision
If robust estimation algorithms like RANSAC are used to remove erroneous corresponding points, then the fundamental matrix accuracy is improved, but the computational time increases
Solution Approach 1:
The patent performs preliminary generation of derived point pairs from detected feature points using geometric transformations, creating an expanded set of corresponding points before applying robust estimation. This preliminary action structures the data in a way that may reduce the impact of erroneous points and potentially reduce the computational burden of subsequent RANSAC processing by providing more reliable correspondence information upfront
Data Source
AI summary
A fundamental matrix generation apparatus performs: detecting three or more feature point pairs from a first image and a second image; detecting, for each of the feature point pairs, a derived point pair that is a pair of a derived point separated by a first distance in a first direction from a point on the first image included in the feature point pair and a derived point separated by a second distance in a second direction from a point on the second image included in the feature point pair; generating a fundamental matrix or the fundamental matrix representing an epipolar constraint on a point on the first image and a point on the second image by using the detected feature point pairs and derived point pairs.