AR Camera Pose Tracking with 2D-2D and 3D-2D Key Points

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing SLAM systems face errors and tracking losses due to insufficient key points during camera pose estimation, particularly in fast movements, leading to inaccurate positioning and navigation.

Innovation Solution

A method combining 3D-2D and 2D-2D correspondences to enhance key point estimation by using previously tracked key points and triangulation, ensuring sufficient key points for accurate camera pose determination.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If only 3D-2D correspondences are used for camera pose estimation, then the estimation process is simpler, but the number of key points is insufficient leading to tracking losses

Engineering Contradiction:
Improvetracking stabilityVSAvoidnumber of key points
Core Design Contradiction:
ReliabilityVSQuantity of substance

Solution Approach 1:

The patent combines 3D-2D correspondences and 2D-2D correspondences into a unified camera pose estimation framework. The 2D-2D correspondence module tracks key points across consecutive image frames, supplementing the key points from 3D-2D triangulation. This merging ensures sufficient key point quantity for reliable tracking while maintaining estimation accuracy through the integrated optimization process.

Inventive Principle:
Principle #5Merging (Combining)

2Reliability

If more key points are used for camera pose estimation, then tracking stability improves, but the computational complexity increases

Engineering Contradiction:
Improvetracking stabilityVSAvoidcomputational complexity
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

The patent divides the camera pose estimation process into separate functional modules: a 3D-2D correspondence module for initial key point generation through triangulation, and a 2D-2D correspondence module for additional key point tracking. This segmentation allows each module to operate independently with optimized algorithms, managing computational complexity while achieving reliable tracking through combined results.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The system performs preliminary key point tracking using 2D-2D correspondences before final camera pose estimation. By pre-identifying and tracking key points across frames, the system prepares sufficient correspondence data in advance, reducing the computational burden during the actual pose estimation phase and improving overall processing efficiency.

Inventive Principle:
Principle #10Preliminary action

3Quantity of substance

If 3D-2D and 2D-2D correspondences are combined, then key point quantity increases, but the estimation process becomes more complex

Engineering Contradiction:
Improvenumber of key pointsVSAvoidestimation process complexity
Core Design Contradiction:
Quantity of substanceVSDevice complexity

Solution Approach 1:

The patent implements a unified optimization framework that handles both 3D-2D and 2D-2D correspondences simultaneously. This multi-functional estimation process uses a single mathematical model to process different types of correspondences, eliminating the need for separate estimation pipelines and reducing overall process complexity despite handling multiple key point sources.

Inventive Principle:
Principle #6Universality (Multi-functionality)

Data Source

PatentEP3664033B1Method for displaying augmented reality and method and device for determining pose information
Publication Date: 2025.11.19 TENCENT TECHNOLOGY (SHENZHEN) CO LTD
  • EP3664033B1 patent drawingFigure 1~2
  • EP3664033B1 patent drawingFigure 3
  • EP3664033B1 patent drawingFigure 4

AI summary

Embodiments of the present application disclose a method for determining pose information of an image capturing device, including: acquiring, by an image capturing device, a first image of a target object; extracting, by the image capturing device, coordinate information of a first two-dimensional (2D) key point corresponding to the target object from the first image; acquiring, by the image capturing device, a second image of the target object; acquiring, by the image capturing device, coordinate information of a second 2D key point in the second image according to the coordinate information of the first 2D key point; acquiring, by the image capturing device, coordinate information of a three-dimensional (3D) key point corresponding to the target object; and determining, by the image capturing device, camera pose information of the second image according to the coordinate information of the first 2D key point, the coordinate information of the second 2D key point, and the coordinate information of the 3D key point. In the embodiments of the present application, more key points can be obtained during the estimation of camera pose information, so that errors in camera pose information are reduced, and a tracking loss because of an excessively small quantity of key points in tracking can be avoided.