AR Camera Orientation Repositioning with Hash-Indexed Keyframes
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing SLAM repositioning methods in augmented reality (AR) scenarios, such as those involving monocular vision, face challenges in maintaining continuous tracking due to feature point loss, leading to poor repositioning success rates, especially when cameras move intensely or towards intense light sources or white walls.
Innovation Solution
A repositioning method that involves dividing the camera movement into tracking processes using anchor images and performing repositioning on current images relative to keyframes stored in a database, utilizing Hash index information to accelerate feature point matching and reduce storage space requirements.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional SLAM repositioning method is used, then repositioning can be performed, but the success probability is poor and tracking is frequently interrupted
Solution Approach 1:
The patent segments the tracking process into multiple stages using anchor images and keyframes. Instead of attempting to track all features simultaneously, it divides the image into regions of interest around anchor points and uses keyframes as intermediate reference points. This segmentation allows the system to handle complex scenes more reliably and maintain continuous tracking even when global feature matching fails.
Solution Approach 2:
The patent introduces keyframes as intermediary elements between anchor images and current images. Keyframes serve as mediator frames that contain pre-extracted and stored feature information, enabling the system to perform repositioning through multiple intermediate steps rather than direct matching. This intermediary approach increases the success probability by providing multiple reference points and reducing the impact of single-point failures.
2Measurement precision
If feature point matching is performed on all images, then tracking accuracy is maintained, but storage space requirements increase significantly
Solution Approach 1:
The patent extracts and stores only the essential feature information from images at keyframes rather than storing complete image data. By extracting feature points, their coordinates, and corresponding descriptor information, the system reduces storage requirements while maintaining tracking accuracy. The keyframe database stores only the necessary feature extraction results, not the full images themselves.
Solution Approach 2:
The patent performs feature extraction and preparation in advance during the anchor image processing stage. By pre-extracting features from anchor images and storing them as keyframes before they are needed for repositioning, the system avoids the need to process all images in real-time during tracking. This preliminary action reduces computational load and storage requirements during the actual tracking process.
3Duration of action of stationary object
If repositioning is performed on current image and first anchor image, then tracking can be maintained, but the process is prone to interruption when feature points are lost
Solution Approach 1:
The patent implements a dynamic keyframe selection mechanism that adapts to changing scene conditions. The system dynamically determines which keyframes to use based on the current image content and the stability of feature matching. When feature points are lost or tracking becomes unstable, the system can switch to alternative keyframes or adjust the repositioning strategy, thereby maintaining tracking duration while improving reliability through adaptive response to environmental changes.
Data Source
Figure 1~2
Figure 3~4
Figure 5~6(3)
AI summary
The present invention relates to the field of augmented reality. Disclosed are a method, device and apparatus for repositioning in a camera orientation tracking process, and a storage medium. The method comprises: acquiring a current image captured after the ith anchor image in a plurality of anchor images; if the current image satisfies a repositioning condition, selecting a target keyframe image from a keyframe image library according to hash index information; performing a second repositioning operation on the current image with respect to the target keyframe image; and performing, according to a positioning result of a first repositioning operation and a positioning result of the second repositioning operation, calculation to obtain a camera orientation parameter of a camera upon capturing the current image. The keyframe image library can comprise multiple keyframe images. Availability of different kayframe images covering the surrounding environment where a camera captures images ensures an overwhelming probability of successful repositioning, thereby increasing a success rate of a repositioning process.