Camera 3D Pose Correction via Georegistered Model
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing augmented reality systems face inaccuracies in georegistering camera poses due to limitations in measurement systems, leading to incorrect superimposition of AR content relative to environmental features, exacerbated by environmental conditions and component degradation.
Innovation Solution
A system and method for determining the corrected 3D pose of a non-georegistered camera using a georegistered camera and a 3D geographic model, by comparing sensor images with virtual images from the model to correct position and orientation inaccuracies.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If measurement systems are used to determine camera position and orientation, then the system can provide location and orientation data, but the accuracy deteriorates due to inherent precision limitations and environmental conditions
Solution Approach 1:
The patent introduces an intermediary georegistration process that mediates between the measurement system and the final AR content placement. Instead of directly using measurement system output, the system georegisters camera images to a 3D geographic model, using the model as an intermediary reference frame to correct pose inaccuracies and improve reliability
Solution Approach 2:
The system implements feedback by continuously comparing measured camera pose with pose derived from georegistration to the 3D geographic model. The discrepancy between these two measurements feeds back into the system to correct ongoing pose estimates, improving accuracy over time despite environmental conditions and component degradation
2Ease of operation
If AR content is superimposed using measured camera pose, then the system can display augmented reality content, but the content alignment deteriorates due to pose inaccuracies
Solution Approach 1:
The 3D geographic model serves as an intermediary reference that mediates between the camera view and the AR content placement. By georegistering the camera image to this model and then placing content in the model's coordinate system, the system achieves precise alignment despite measurement inaccuracies
Solution Approach 2:
The system creates a copied or replicated version of the real-world scene in the 3D geographic model. AR content is placed in this copied representation, which then maps back to the real world with corrected positioning, improving alignment precision while maintaining ease of operation
Data Source
Figure 1
Figure 2
Figure 3A~3B
AI summary
Method/system for determining pose of camera using another camera imaging common scene, by capturing first scene image with first camera and obtaining first camera pose by georegistering to 3D-model, extracting scene features in first image, determining 3D coordinates of extracted features by mapping to 3D model, transmitting feature descriptors and feature coordinates, capturing second scene image with second camera, extracting scene features in second image and matching with first image features, determining second camera pose using 3D scene coordinates and corresponding 2D projections in second image. Method/system for determining pose of one camera using another camera in same camera assembly with known relative pose between cameras in assembly, by capturing image of scene with first camera and determining 3D coordinates of first camera by georegistering to 3D-model, and determining pose of second camera based on determined global pose of first camera and relative pose between second camera and first camera.