Stereo Vision Correspondence Matching for Non-Rigid Object Reconstruction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current stereo vision technologies face challenges in accurately identifying and recognizing non-rigid, amorphous objects such as wire harnesses due to their variable shapes and appearances, which complicates their application in industrial settings like assembly lines where precise object manipulation is required.
Innovation Solution
A method is developed to identify corresponding objects in stereoscopic images by determining connected components, feature descriptors, and applying global and local homography to match and align features, ensuring accurate correspondence and 3D reconstruction of objects like wire harnesses.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If traditional shape-based recognition methods are used, then simple rigid objects can be identified, but non-rigid objects with variable shapes like wire harnesses cannot be accurately recognized
Solution Approach 1:
The patent segments the wire harness into multiple connected components, where each component represents a distinct segment of the harness. This segmentation allows the system to handle the variability of non-rigid objects by treating them as compositions of simpler, more recognizable parts, thereby improving both adaptability and measurement precision
Solution Approach 2:
The patent transitions from 2D image analysis to 3D reconstruction by using stereo vision. By capturing images from multiple viewpoints and reconstructing the wire harness in three dimensions, the system can accurately identify non-rigid objects despite their variable shapes, resolving the contradiction between adaptability and measurement precision
2Measurement precision
If stereo vision is applied to non-rigid objects, then 3D reconstruction capability is improved, but correspondence matching between images becomes more difficult
Solution Approach 1:
The patent divides the wire harness into connected components in each stereo image, then performs correspondence matching between these segmented components rather than attempting to match the entire variable-shape object. This segmentation approach significantly reduces the difficulty of feature correspondence matching while maintaining 3D reconstruction accuracy
Solution Approach 2:
The patent applies different processing strategies to different parts of the wire harness based on their local characteristics. By treating each connected component with localized matching algorithms appropriate to its specific geometry, the system overcomes the general difficulty of matching non-rigid objects in stereo vision
3Adaptability or versatility
If wire harnesses are used in assembly lines, then manufacturing flexibility is improved, but automated manipulation becomes more challenging due to shape variability
Solution Approach 1:
The patent uses stereo vision to reconstruct the wire harness in 3D space, providing robots with depth information and spatial understanding. This 3D reconstruction enables automated manipulation systems to grasp and handle variable-shape wire harnesses effectively, maintaining manufacturing flexibility while improving ease of automated operation
Solution Approach 2:
The patent implements a feedback loop where the vision system continuously identifies and tracks the wire harness position and orientation, then provides this information to the manipulation system. This real-time feedback enables adaptive automated manipulation that can handle shape variability while maintaining manufacturing flexibility
Data Source
AI summary
In a stereoscopic pair of images, global homography at the image level is applied to feature points extracted from connected components (CC) to identify corresponding CC's and feature points, and to discard any CC's that do not have a corresponding pair in the stereoscopic pair of images. Local homography at the CC level is then applied to individual footprint areas of the previously identified paired CC to further clean feature point correspondence. Any CC or feature point or pixel within a paired CC footprint not satisfying local homography constraint is discarded. A correspondence is also extrapolated between unknown pixels within a paired CC footprint using a weighing mechanism and the unknown pixel's surrounding pixels that do have a known correspondence. This provides a dense correspondence of pixels, or feature points, which is then used to create a dense 3D point cloud of identified objects within a 3D space.