AR Camera Pose Relocalization via Anchor Feature Filtering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current SLAM relocalization methods in augmented reality (AR) are prone to tracking interruptions and accumulate errors due to the loss of feature points, especially in scenarios with intense camera movement or complex environments, making them unsuitable for anchor-SLAM algorithms.
Innovation Solution
A relocalization method that filters matching feature point pairs based on constraint conditions, such as uniqueness, Epipolar constraints, and regional representation, to reduce the number of feature points required for calculation and improve matching accuracy, allowing for accurate camera pose tracking by switching between anchor images when tracking fails.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If feature point tracking is performed between current image and anchor image to obtain camera pose, then camera position tracking is achieved, but tracking interruptions occur when feature points are lost
Solution Approach 1:
The system performs relocalization by matching current image features with pre-stored anchor image features before complete tracking failure occurs. This preliminary action of maintaining a feature database from anchor images enables recovery when tracking is interrupted, preventing feature point loss from causing complete tracking failure.
2Reliability
If SLAM relocalization method is used when feature points are lost, then relocalization can be performed, but the method is not applicable to anchor-SLAM algorithms and accumulated errors cannot be eliminated
Solution Approach 1:
The patent extracts and isolates the relocalization function as a separate module that specifically operates between current images and anchor images. This extracted relocalization module is integrated into the anchor-SLAM algorithm, enabling independent optimization of relocalization accuracy without affecting the main tracking pipeline, thus eliminating accumulated errors while maintaining algorithm applicability.
Solution Approach 2:
Anchor images serve as an intermediary between the initial reference frame and current images. The relocalization process uses anchor image features as a mediator to recover tracking when feature points are lost, providing a bridge that maintains measurement precision while enabling relocalization capability in anchor-SLAM frameworks.
3Measurement precision
If multiple feature point pairs are used for calculation to ensure accuracy, then matching precision is improved, but matching speed decreases
Solution Approach 1:
The system applies different processing qualities to different parts of the feature matching process. High-precision feature extraction and matching is performed only on anchor images and their corresponding current images, while other tracking operations use standard processing. This localized high-quality processing ensures matching accuracy where critical while maintaining overall system speed.
Data Source
Figure 1~2
Figure 3~4
Figure 5~6
AI summary
This application discloses a relocalization method and apparatus in a camera pose tracking process, a device, and a storage medium, belonging to the field of augmented reality (AR). The method includes: obtaining a current image acquired after an ith anchor image in a plurality of anchor images; obtaining an initial feature point and an initial pose parameter in the first anchor image in the plurality of anchor images in a case that the current image satisfies a relocalization condition; performing feature point tracking on the current image relative to the first anchor image, to obtain a plurality of matching feature point pairs; filtering the plurality of matching feature point pairs according to a constraint condition, to obtain a filtered matching feature point pair; calculating a pose change amount of a change of a camera from the initial pose parameter to a target pose parameter according to the filtered matching feature point pair; and performing relocalization according to the initial pose parameter and the pose change amount to obtain the target pose parameter of the camera.