Edge-based 3D AR Tracking via Grouped Edgelet Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current edge-based three-dimensional tracking and registration methods for augmented reality face challenges in complex environments due to low robustness and high computational complexity, with existing methods either being fast but not robust or robust but slow.
Innovation Solution
The method involves extracting edgelet features of a three-dimensional object, dividing them into groups based on geometric structure, and performing searches along average normal vectors to find matching points, allowing for robust and fast calculation of the object's pose using the RANSAC and IRLS algorithms.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If edge matching method is used to calculate precise pose, then calculation speed is fast, but robustness deteriorates in complex environments
Solution Approach 1:
The patent segments the edgelet features into multiple groups based on their spatial distribution and geometric relationships. By dividing the feature set into manageable groups, the system can perform matching operations more efficiently while maintaining robustness through group-level validation and aggregation of matching results.
Solution Approach 2:
The patent transforms the matching problem by changing parameters from traditional edge point matching to edgelet feature matching with grouped processing. It introduces new parameters such as group coherence metrics and geometric constraints that enhance robustness while maintaining computational efficiency through optimized search strategies.
2Reliability
If deep learning method is used to calculate precise pose, then robustness is improved, but calculation speed deteriorates
Solution Approach 1:
The patent applies segmentation by dividing edgelet features into groups that can be processed independently and in parallel. This reduces the computational burden compared to holistic deep learning approaches while maintaining robustness through the collective information from multiple feature groups.
Solution Approach 2:
The patent substitutes complex deep learning mechanical systems with a more efficient edgelet-based geometric constraint system. By using explicit geometric relationships and algebraic optimization instead of neural network inference, it achieves comparable robustness with significantly improved calculation speed.
3Measurement precision
If traditional edge matching is performed on all edgelet features, then matching completeness is improved, but computational complexity increases
Solution Approach 1:
The patent segments the comprehensive matching task into group-level operations. By processing features in groups and using group-level constraints, it achieves matching completeness equivalent to individual feature matching while reducing computational complexity through the aggregation effect and reduced search space.
Solution Approach 2:
The patent applies partial action by performing matching operations on representative groups of features rather than exhaustively processing every individual feature. This selective approach maintains matching completeness for critical features while reducing overall computational complexity through intelligent sampling and grouping strategies.
Data Source
AI summary
The present disclosure discloses an edge-based three-dimensional tracking and registration method and apparatus for augmented reality, and a storage medium. The method includes: obtaining an initial pose of a three-dimensional object in an image; extracting edgelet features of a contour of the three-dimensional object based on the initial pose; dividing the edgelet features into groups, and performing searching on the image to find a matching point corresponding to respective edgelet feature in each group; and calculating a pose of the three-dimensional object based on matching points found in the groups.