Camera Pose Estimation from Sparse 3D Maps in GPS-Denied Garages
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing vehicle localization systems, such as GPS and INS, are inadequate for providing accurate, real-time positional information in GPS-denied environments like parking garages, where automated parking and semi-autonomous operations are necessary, due to signal reception issues and positional drift.
Innovation Solution
A camera-based localization scheme that uses a map created through a teach and repeat process, where on-vehicle cameras match features against pre-built maps to estimate vehicle pose, utilizing image feature points, map feature descriptors, and reprojection error to determine the vehicle's location within a threshold value.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If GPS and INS systems are used for vehicle localization, then positional information can be obtained in open environments, but accuracy deteriorates in GPS-denied environments like parking garages
Solution Approach 1:
The patent introduces a camera-based visual localization system as an intermediary between the vehicle and the environment, using image feature matching to determine position when GPS is unavailable. The camera captures images, extracts feature points, and matches them against pre-stored map data to calculate vehicle location and orientation, serving as a mediator that enables localization in GPS-denied environments.
Solution Approach 2:
The patent segments the localization problem into distinct operational modes: GPS-available mode and GPS-denied mode. In GPS-denied mode, it further segments the visual localization task into image acquisition, feature point extraction, feature matching, and pose calculation steps, allowing the system to adaptively switch between different localization methods based on environmental conditions.
2Adaptability or versatility
If camera-based localization is implemented, then localization works in GPS-denied environments, but system complexity increases
Solution Approach 1:
The patent makes the camera system multi-functional by using it for both primary localization (via feature matching) and as a backup to GPS. The same camera hardware and processing pipeline serve multiple purposes: capturing images for visual localization, providing environmental awareness, and enabling operation in diverse conditions including tunnels and parking garages, thereby reducing the need for separate dedicated systems.
Solution Approach 2:
The patent performs preliminary actions by pre-processing and storing three-dimensional map data and feature point information before the vehicle needs localization. The map data is prepared in advance with known feature points and descriptors, allowing the real-time localization system to simply compare current camera images against this pre-computed reference data, reducing computational complexity during actual localization operations.
3Measurement precision
If real-time localization is achieved through feature matching, then positional accuracy is improved, but processing time increases
Solution Approach 1:
The patent applies partial action by not requiring complete matching of all feature points in the image. Instead, it identifies and matches a sufficient subset of key feature points to achieve accurate localization. The system extracts multiple feature points but only needs to match enough of them to confidently determine vehicle pose, reducing processing time while maintaining accuracy.
Solution Approach 2:
The patent performs preliminary computation by pre-calculating and storing three-dimensional map data, feature point locations, and descriptors before the vehicle arrives at the location. This pre-processing allows the real-time system to simply retrieve and compare data rather than computing everything from scratch, significantly reducing processing time during actual localization while maintaining high accuracy through comprehensive feature matching.
Data Source
AI summary
Techniques for localizing a vehicle include obtaining an image from a camera, identifying a set of image feature points in the image, obtaining an approximate location of the vehicle, determining a set of sub-volumes (SVs) of a map to access based on the approximate location, obtaining map feature points and associated map feature descriptors associated with the set of SVs, determining a set of candidate matches between the set of image feature points and the obtained map feature points, determining a set of potential poses of the camera from candidate matches from the set of candidate matches and an associated reprojection error estimated for remaining points to select a first pose of the set of potential poses having a lowest associated reprojection error, determining the first pose is within a threshold value of an expected vehicle location, and outputting a vehicle location based on the first pose.