3D Map Pose Estimation With Key Frame Selection in GPS-Shaded Areas
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing GPS-based location measurement methods fail in GPS shaded areas, and camera-based positioning methods require large data capacity and significant computational resources.
Innovation Solution
A method for generating a 3D map using key frames from captured images and determining pose information by selecting matching frames based on valid pose prediction, employing local and global feature matching with neural networks to reduce data capacity and computational load.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If camera-based positioning method is used, then positioning accuracy is improved, but data capacity and computational load increase
Solution Approach 1:
The patent segments the complete image data into key frames selected at specific intervals based on pose prediction validity. Only these key frames are stored in the 3D map, while other frames use predictive pose estimation. This segmentation reduces data capacity requirements while maintaining positioning accuracy through selective detailed storage and predictive interpolation for intermediate states.
2Measurement precision
If camera-based positioning method is used, then positioning accuracy is improved, but computational load increases
Solution Approach 1:
The patent implements a dynamic computational strategy where the processing intensity varies based on the validity of pose predictions. When pose prediction is valid, minimal computational effort is required (only checking prediction validity). When prediction is invalid, full computational processing is performed to select new key frames. This dynamic approach reduces average computational load while maintaining positioning accuracy.
Solution Approach 2:
The patent performs preliminary pose prediction using previous frame information before full image processing. This preliminary action allows the system to anticipate the pose state and determine whether full computational processing is necessary, thereby reducing overall computational load while maintaining positioning accuracy through predictive preprocessing.
3Quantity of substance
If key frames are selected from captured images, then 3D map data capacity is reduced, but system complexity increases
Solution Approach 1:
The patent uses feedback from pose prediction validity to control key frame selection. The system continuously monitors whether pose predictions are valid and uses this feedback to determine when to select new key frames. This feedback mechanism automates the key frame selection process, reducing the need for complex manual configuration while maintaining reduced data capacity through intelligent, adaptive sampling.
Data Source
AI summary
In accordance with an aspect of the present disclosure, there is provided a method of determining a pose of a camera included in a user terminal. The method comprises, based on whether a first pose of the camera is validly predicted in a first query image in which a location of the user terminal is captured by the user terminal, determining a prediction mode for predicting a second pose of the camera in a second query image in which the location is captured by the user terminal after the first query image; and determining the second pose based on the determined prediction mode.