Structured Light Matching Using Epipolar Constraints
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current three-dimensional scanning technologies using structured light stereovision face challenges in efficiently matching pixels between images from two cameras, especially when projecting complex patterns, which increases ambiguity and reduces precision, particularly in portable scanners where all necessary information must be found within a single frame without additional constraints on the scene.
Innovation Solution
A method that extracts blobs from image reflections, labels them uniquely, selects epipolar planes, identifies plausible combinations, calculates matching errors, and determines the most probable combination using a figure of merit to match points between images, allowing for efficient matching within a single frame without requiring additional constraints on the scene or hardware complexity.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If the number of projected stripes is increased to reduce scanning time, then productivity is improved, but the complexity of matching image points grows exponentially and introduces ambiguities
Solution Approach 1:
The patent segments the matching problem by introducing epipolar constraints that divide the image space into manageable epipolar lines. Instead of matching points across the entire image, the algorithm restricts search to specific epipolar lines, reducing the complexity from O(N^2) to O(N) where N is the number of stripes. This segmentation allows handling of complex patterns with many stripes while maintaining efficient matching.
Solution Approach 2:
The patent introduces epipolar geometry as an intermediary constraint between the projected pattern and image matching. The epipolar lines act as intermediaries that guide the matching process, providing a geometric framework that reduces search space and eliminates ambiguities. This intermediary structure enables efficient matching even with increased stripe density.
2Measurement precision
If a grid of stripes is projected to enable surface reconstruction, then measurement precision is improved, but the complexity of matching image points grows exponentially
Solution Approach 1:
The patent applies segmentation by dividing the grid pattern matching into independent epipolar line problems. Each epipolar line contains a subset of grid intersections, and matching is performed independently along each line. This segmentation reduces the overall complexity from exponential to linear, while preserving the precision benefits of grid-based surface reconstruction.
Solution Approach 2:
The patent transitions from two-dimensional image space matching to a constrained one-dimensional matching along epipolar lines. By projecting the 3D geometric constraints into 2D image space as linear epipolar constraints, the problem dimensionality is effectively reduced, making grid pattern matching computationally feasible while maintaining precision.
3Measurement precision
If additional cameras are added to resolve ambiguities, then measurement precision is improved, but device complexity increases and frame rate limit is reduced
Solution Approach 1:
The patent uses epipolar geometry as a mathematical intermediary that resolves ambiguities without additional hardware. The epipolar constraints provide sufficient geometric information to disambiguate matching points, replacing the need for additional cameras while maintaining precision. This intermediary approach achieves the same effect as multiple views with minimal hardware.
Solution Approach 2:
The patent changes the parameter space by introducing epipolar angle and distance parameters that uniquely identify matching points. Instead of relying on multiple camera views, the system uses parametric epipolar constraints to resolve ambiguities, achieving precise matching with the original two-camera configuration.
4Device complexity
If epipolar constraints are used to limit search to a single line, then matching complexity is reduced, but ambiguities may still exist in complex patterns
Solution Approach 1:
The patent implements feedback by using the known projected pattern geometry to validate and refine matching results. The system compares the observed epipolar line intersections with the expected pattern geometry, providing feedback that resolves ambiguities. This feedback mechanism ensures accurate matching even in complex patterns by continuously verifying results against the known pattern structure.
Solution Approach 2:
The patent performs preliminary action by pre-calculating and storing the expected epipolar line intersections for the known projected pattern. Before performing actual matching, the system prepares the reference epipolar geometry, which enables rapid and unambiguous identification of matching points during the scanning process. This preliminary preparation eliminates ambiguities by providing reference information in advance.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
This approach enables precise matching of points in each frame, reducing ambiguity and maintaining precision while avoiding the need for complex projectors or additional constraints, thus enhancing the flexibility and accuracy of three-dimensional scanning with two cameras.
Implementation Method 1
a light projector unit projecting sheets of light onto the scene thereby providing a projected pattern on the scene
Implementation Method 2
the images each containing a reflection of the projected pattern on the scene
Data Source
AI summary
A method for matching points between two images of a scene acquired by two cameras, the method including: extracting blobs from the reflection in the images and individually labelling the blobs with a unique identifier; selecting a selected epipolar plane and defining a conjugate epipolar line on each of the images; identifying plausible combinations including a light sheet label of the projected pattern and the unique identifier for a plausible blob selected from the blobs in each of the images; calculating a matching error for each of the plausible combinations; repeating the steps of selecting, identifying and calculating for at least some epipolar planes; determining a most probable combination by computing a figure of merit for the plausible combinations; identifying matching points in the frame from the most probable combination.