Visual Localization Map Update via Consecutive Image Pose Estimation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing map generation technologies for mobile robots in new environments are inefficient in updating maps to account for changes in space and state, particularly in verifying the position of mobile devices based on images, and do not effectively utilize consecutive query information for location-based services.
Innovation Solution
A method and system that calculates the relative pose relationship between consecutive images, estimates the absolute pose of each image, and updates the visual localization map using a graph structure to incorporate images with failed pose calculations, ensuring the map remains up-to-date and accurate for pose estimation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If traditional map generation technology is used for mobile robots in new environments, then the robot can generate a grid map based on distance information, but the map cannot be efficiently updated to account for changes in space and state
Solution Approach 1:
The system dynamically updates the visual localization map by incorporating new images and their absolute poses as they become available, rather than using static map generation. The map structure is continuously adapted to reflect changes in the environment, allowing the system to maintain accuracy in dynamic settings while efficiently processing incoming data streams.
Solution Approach 2:
The system changes the parameters used for map representation from traditional grid-based distance information to visual features and absolute pose information. By transforming the map into a visual localization map with different parameter structures, the system achieves both better adaptability to environmental changes and improved update efficiency through graph-based optimization.
2Measurement precision
If visual localization map updating is performed using only successful pose calculations, then the map accuracy is maintained, but images with failed pose calculations are discarded causing information loss
Solution Approach 1:
The system converts the previously harmful effect of pose calculation failures into a beneficial opportunity. Instead of discarding images that fail pose calculation, the system incorporates them into the map update process by estimating their absolute poses through graph optimization. This transforms data loss into additional useful information that enhances map coverage and accuracy.
Solution Approach 2:
The system introduces graph optimization as an intermediary mechanism between successful pose calculations and map updates. This intermediary process estimates absolute poses for images that directly fail pose calculation, serving as a bridge that preserves and utilizes information from previously discarded images while maintaining overall map accuracy.
3Ease of operation
If the system processes only consecutive query information from electronic devices, then location-based services are enabled, but the map update process becomes complex requiring graph structure optimization
Solution Approach 1:
The system segments the map update process into distinct functional modules: receiving consecutive images, calculating relative poses, determining absolute poses through graph optimization, and updating the visual localization map. This segmentation allows each module to handle specific tasks independently, managing complexity while enabling continuous location-based services through streamlined processing pipelines.
Data Source
AI summary
A map updating method may include calculating a relative pose relationship between consecutive images; calculating an absolute pose of each image based on the relative pose relationship between the consecutive images; and updating a map used for visual localization (VL) based on the absolute pose.