Camera Pose Determination with Adaptive Feature Point Tracking
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing AR technologies struggle to accurately determine the position and attitude of a camera when there are excessive changes, leading to failure in feature point tracking.
Innovation Solution
A method that extracts and tracks multiple feature points, including a second feature point when the first fails, using grid areas for spatial dispersion and homography matrix decomposition to stabilize the tracking process without requiring a pre-set marked image.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a single feature point is tracked in the marker image, then the tracking process is simple and fast, but the tracking fails when the camera position or attitude changes excessively
Solution Approach 1:
The marker image is divided into multiple grid areas, and each grid area contains a feature point. This segmentation ensures that when the camera position or attitude changes, at least one feature point remains visible and trackable, preventing complete tracking failure.
Solution Approach 2:
The patent transitions from tracking a single feature point (0-dimensional) to tracking multiple feature points distributed across 2-dimensional grid areas. This dimensional expansion provides redundancy, ensuring that if some feature points become invisible due to camera movement, others remain visible and can maintain tracking.
2Reliability
If multiple feature points are extracted and tracked, then the tracking robustness improves, but the computational complexity increases
Solution Approach 1:
The marker image is segmented into grid areas with feature points distributed across them. This structured segmentation simplifies the extraction process by providing clear spatial regions to search for feature points, reducing the complexity of finding multiple feature points while maintaining robustness.
Solution Approach 2:
The patent uses parameter changes in the form of grid area divisions and feature point distribution patterns. By changing the spatial parameters of feature point locations rather than using complex extraction algorithms, the system achieves robust tracking with controlled computational complexity.
3Measurement precision
If a pre-set marked image is used, then the tracking accuracy is high, but the system requires presetting which limits adaptability
Solution Approach 1:
The system performs preliminary actions by capturing a reference image and extracting feature points from it before actual tracking begins. This preliminary feature point extraction from the captured image itself (rather than a pre-set marker) maintains tracking accuracy while removing the limitation of requiring pre-configured marker images, thereby expanding adaptability.
Solution Approach 2:
The system uses the captured image itself to generate the feature points for tracking, rather than requiring an external pre-set marker. The captured image serves双重 purposes: as the scene to be tracked and as the source of tracking features, making the system self-sufficient and adaptable to any scene without preset requirements.
Data Source
Figure 1
Figure 2
Figure 3~4
AI summary
The present invention relates to the technical field of computers. Disclosed are a pose determination method and device, an intelligent apparatus, and a storage medium. The method comprises: in the event of no preset anchor image, an intelligent apparatus acquiring an image captured by a camera, and setting the captured image as an anchor image (301); acquiring, by means of tracking a first feature point, a pose parameter of a first image captured by the camera with respect to the anchor image (302); if the first image does not satisfy a feature point tracking condition, extracting a second feature point from the first image (303); and acquiring, by means of tracking the first feature point and the second feature point, a pose parameter of a second image captured by the camera with respect to the anchor image, and determining a pose of the camera according to the pose parameter (304). In the invention, when an image does not satisfy a feature point tracking condition, a new feature point is extracted, thereby preventing the problem in which a feature point is unable to be tracked due to too many changes in a position or an orientation of a camera, enhancing robustness, and increasing tracking precision of the camera.