Visual Localization Robustness via Depth Map Verification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing visual localization systems suffer from weak localization robustness due to their high dependence on visual algorithms, leading to incorrect posture estimation and localization errors in complex real-world scenarios.
Innovation Solution
A method is introduced that adds a parallel processing module to the existing visual localization system. This module uses depth maps, point cloud maps, and relocation postures to obtain three-dimensional coordinates of spatial obstacles and match them with environmental coordinates, thereby determining the correctness of relocation postures and improving localization robustness.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If visual algorithms are used for localization, then the system can achieve localization function, but the localization robustness becomes weak due to high dependence on visual algorithms
Solution Approach 1:
The patent introduces depth map information as an intermediary element between the visual algorithm and the relocation verification process. The depth map provides additional spatial information that mediates the verification of relocation postures, reducing direct dependence on visual algorithms alone while maintaining system functionality.
Solution Approach 2:
The localization system is segmented into multiple independent modules: visual algorithm module, depth map generation module, and relocation verification module. This segmentation allows each module to operate independently, reducing the system's dependence on any single algorithm while improving overall robustness through modular verification.
2Measurement precision
If additional sensors are added to improve localization accuracy, then localization precision improves, but device complexity and cost increase
Solution Approach 1:
The existing camera hardware is made multi-functional by utilizing it not only for visual recognition but also for generating depth maps through stereo vision or focus variation techniques. This eliminates the need for additional depth sensors while improving localization accuracy through multi-functional use of existing components.
Solution Approach 2:
Instead of adding physical depth sensors, the system creates a virtual copy of spatial information through depth map generation from visual data. This virtual representation serves the same purpose as physical depth sensors would, but without the associated hardware complexity and cost.
3Device complexity
If relocation module heavily depends on visual algorithm, then the system structure remains simple, but localization errors occur in complex real-world scenarios
Solution Approach 1:
The system performs preliminary generation of depth map information and pre-computation of expected spatial relationships before the relocation verification step. This preliminary action provides additional verification data that helps detect localization errors in complex scenarios without requiring complex system restructuring.
Solution Approach 2:
The patent implements a feedback mechanism where the depth map information and spatial obstacle point verification provide feedback to the relocation module. This feedback loop allows the system to detect and correct localization errors by comparing expected spatial relationships with actual observations, improving reliability while maintaining relatively simple system structure.
Data Source
Figure 1
Figure 2
Figure 3~4
AI summary
A method and device for obtaining localization information and a medium are provided. The method includes: obtaining image information and related information of the image information, wherein the related information includes a depth map, a point cloud map, relocation postures and relocation variance after relocation; obtaining three-dimensional coordinates of spatial obstacle points based on the depth map; obtaining target postures and environmental three-dimensional coordinates corresponding to each of the target postures based on the relocation postures, the relocation variance and the point cloud map; scanning and matching the three-dimensional coordinates of the spatial obstacle points with the environmental three-dimensional coordinates to obtain matching result information; and obtaining localization information based on the relocation postures and the relocation variance when the matching result information satisfies a predetermined condition. The method solves the problem of weak localization robustness caused by high dependence of the relocation module on the visual algorithm in the existing visual localization system, and improves the localization robustness.