3D Object Pose Estimation Using Graph Matching for Weak Texture
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for three-dimensional object pose estimation are ineffective for objects with weak texture or repeating texture due to feature point mismatch issues, leading to incorrect pose estimation results.
Innovation Solution
A method involving graph matching algorithms to calculate vertex correspondence between frames, combined with the PnP algorithm, to improve feature point matching accuracy and pose estimation accuracy for three-dimensional objects with weak or repeating textures.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If feature point matching combined with PnP algorithm is used for pose estimation, then the method works well for objects with strong texture, but the matching accuracy deteriorates for objects with weak texture or repeating texture
Solution Approach 1:
The patent introduces a graph matching algorithm as an intermediary step between traditional feature point matching and pose estimation. The graph matching algorithm processes the feature points and their relationships to generate a correspondence relationship that serves as a mediator, filtering out incorrect matches before they reach the PnP algorithm. This intermediary processing layer resolves the contradiction by improving matching reliability for challenging textures while maintaining the overall pose estimation workflow.
Solution Approach 2:
The patent transitions from traditional two-dimensional feature point coordinates to a graph-based representation that incorporates relational information between feature points. By constructing graphs where nodes represent feature points and edges represent spatial relationships, the method adds a dimensional layer of structural information. This graph-based dimension enhances the discriminative power for matching feature points on objects with weak or repeating textures, resolving the adaptability issue.
2Reliability
If RANSAC algorithm is used to remove incorrect matching points, then some incorrect points are eliminated, but the pose estimation accuracy remains insufficient when incorrect points outnumber correct points
Solution Approach 1:
The patent applies graph matching algorithm as a preliminary action before the final pose estimation. By pre-processing the feature point correspondences through graph matching to establish a reliable correspondence relationship, the method proactively eliminates incorrect matches before they can significantly impact the pose estimation. This preliminary filtering action ensures that the subsequent PnP algorithm receives high-quality input, resolving the contradiction between robustness and precision.
Solution Approach 2:
The patent replaces the traditional RANSAC iterative outlier rejection mechanism with a graph-based matching approach. Instead of relying on statistical sampling and iterative refinement, the graph matching algorithm uses structural relationship analysis to directly identify and eliminate incorrect correspondences. This substitution of the matching mechanism provides more reliable filtering of incorrect points, improving both robustness and final pose estimation accuracy.
Data Source
AI summary
The present application provides a method and an apparatus for three-dimensional object pose estimation, a device and a storage medium. The method includes: calculating a graph of a previous frame and a graph of a current frame for a target three-dimensional object; performing a matching calculation on the graph of the previous frame and the graph of the current frame using a graph matching algorithm to obtain a vertex correspondence relationship between the graph of the previous frame and the graph of the current frame; calculating a pose of the target three-dimensional object in the current frame according to the vertex correspondence relationship, a pose of the target three-dimensional object in the previous frame and a PnP algorithm. The matching accuracy of feature points is effectively improved, and thereby the accuracy of three-dimensional object pose estimation is improved.