Stereo Camera Localization Under GNSS Interference
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing VPR techniques for vehicle localization require extensive computing resources and training data, especially when operating in environments with limited field of view and temporary GNSS interference, such as traveling in the opposite direction of the reference database acquisition.
Innovation Solution
Utilizing limited field of view stereo cameras and a method that generates keyframe point clouds, determines similar and opposing viewpoint query matrices, and performs sequence matching to locate vehicles relative to a georeferenced reference database without prior knowledge of the reference viewpoint, reducing computational complexity and resource requirements.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If existing VPR techniques are used for vehicle localization, then localization capability is achieved, but computing resources and training data requirements become excessive
Solution Approach 1:
The patent segments the scene into multiple depth planes (first depth plane, second depth plane, etc.) and processes each plane separately. This segmentation allows the system to handle complex 3D spatial relationships through simpler 2D image processing on each plane, reducing overall computational complexity while maintaining localization accuracy
Solution Approach 2:
The patent creates virtual images by copying and transforming real images through geometric transformations. Virtual images are generated for different depth planes and viewpoints, enabling the system to analyze multiple perspectives without requiring additional physical cameras or extensive training data
2Measurement precision
If existing VPR techniques are used for vehicle localization, then localization capability is achieved, but training data requirements become excessive
Solution Approach 1:
The system performs self-calibration by using geometric relationships between virtual images and real images to automatically determine camera parameters and depth plane configurations. This self-service approach eliminates the need for extensive manual training data collection and processing
Solution Approach 2:
The patent performs preliminary geometric transformations to generate virtual images and establish depth planes before actual localization processing. This preliminary action creates a structured framework that simplifies subsequent localization tasks and reduces the need for extensive training data
3Device complexity
If limited field of view cameras are used, then device complexity is reduced, but localization accuracy in opposing directions deteriorates
Solution Approach 1:
The patent transitions from 2D image analysis to 3D spatial reasoning by introducing depth planes and performing geometric transformations across multiple dimensions. This dimensional expansion allows limited field of view cameras to achieve accurate localization by synthesizing 3D spatial relationships from 2D images
Solution Approach 2:
Virtual images serve as intermediaries between real images captured by limited field of view cameras and the final localization result. These virtual images are generated through geometric transformations and used to bridge the gap between limited camera views and comprehensive spatial understanding
4Measurement precision
If GNSS is used for vehicle localization, then localization accuracy is maintained, but operational reliability deteriorates under temporary interference
Solution Approach 1:
The system prepares alternative localization methods (visual place recognition with depth planes) in advance to cushion against potential GNSS failures. This prior preparation ensures continuous operational reliability by having backup localization capabilities ready before interference occurs
Data Source
AI summary
A computer that includes a processor and a memory, the memory including instructions executable by the processor to generate a current keyframe point cloud based on pairs of stereo images while a stereo camera travels through a scene to determine a similar viewpoint query matrix and an opposing viewpoint query matrix based the current keyframe point cloud. A distance matrix and an opposing view distance matrix can be generated by comparing the similar viewpoint query matrix and the opposing viewpoint query matrix to reference matrices. A relative pose between a stereo camera and a reference can be determined to determine a location in the scene during travel of the stereo camera through the scene by performing sequence matching in the distance matrix and the opposing view distance matrix to determine a minimum sequence.