AR Camera Pose Repositioning via Anchor Feature Point Projection

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing SLAM repositioning methods in augmented reality (AR) have a poor repositioning effect due to accumulated errors and tracking interruptions when feature points are lost, leading to suboptimal camera pose tracking.

Innovation Solution

A repositioning method and apparatus that utilize an inertial measurement unit (IMU) to perform rotation, translation, and projection of initial feature points, searching for target feature points within a first and then a larger range if initial matches are not found, ensuring accurate camera pose recalibration.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If feature point tracking is performed between current image and anchor image to obtain camera pose, then camera pose tracking is achieved, but tracking interruption occurs when feature points are lost

Engineering Contradiction:
Improvetracking continuityVSAvoidfeature point loss
Core Design Contradiction:
ReliabilityVSLoss of information

Solution Approach 1:

The system pre-processes the anchor image to extract and store feature points before tracking begins. When repositioning is needed, these pre-extracted feature points are reused to quickly establish correspondence without requiring full re-tracking, thus maintaining tracking continuity despite feature point loss in the current image.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The patent creates a copy of feature points from the anchor image and uses this copied feature point set for repositioning calculations. This allows the system to perform pose estimation using the anchor image's feature points as a reference, bypassing the need to track lost feature points in the current image.

Inventive Principle:
Principle #26Copying

2Measurement precision

If SLAM repositioning is performed using related art methods, then repositioning is achieved, but accumulated errors are not eliminated leading to poor repositioning effect

Engineering Contradiction:
Improverepositioning accuracyVSAvoidaccumulated error
Core Design Contradiction:
Measurement precisionVSLoss of information

Solution Approach 1:

Instead of tracking from the current image forward to find the next anchor image, the patent inverts the approach by using the anchor image as the reference and repositioning the current image relative to it. This backward repositioning approach eliminates accumulated errors by resetting the reference frame to the known accurate anchor image pose.

Inventive Principle:
Principle #13The other way round (Inversion)

Solution Approach 2:

The anchor image is pre-processed to extract feature points and establish a reference coordinate system before repositioning is needed. This preliminary preparation allows the system to quickly and accurately reposition the current image by comparing against this pre-established reference, eliminating the need to carry forward accumulated errors from previous tracking steps.

Inventive Principle:
Principle #10Preliminary action

Data Source

PatentEP3779883B1Method and device for repositioning in camera orientation tracking process, and storage medium
Publication Date: 2024.10.16 TENCENT TECHNOLOGY (SHENZHEN) CO LTD
  • EP3779883B1 patent drawingFigure 1~2
  • EP3779883B1 patent drawingFigure 3~4
  • EP3779883B1 patent drawingFigure 5~6

AI summary

A method and device for repositioning in a camera orientation tracking process and a storage medium, relating to the technical field of augmented display. The method comprises: acquiring a current image captured after the ith anchor image in a plurality of anchor images (501); if the current image satisfies a repositioning condition, acquiring an initial feature point and an initial pose parameter of the first anchor image in the plurality of anchor images, the initial pose parameter indicating an orientation of a camera when capturing the first anchor image (502); performing feature point tracking on the current image, with respect to the first anchor image, to obtain a target feature point matching the initial feature point (503); performing, according to the initial feature point and the target feature point, calculation to obtain a pose variation of the camera when changing from a first camera orientation to a target camera orientation, wherein the target camera orientation is a camera orientation in which the camera captures the current image (504); and performing, according to the initial pose parameter and the pose variation, repositioning to obtain a target pose parameter corresponding to the target camera orientation (505). The method realizes repositioning in an Anchor-SLAM algorithm used to perform tracking on the basis of multiple continuous anchor images, thereby reducing the possibility of an interrupted tracking process.