Camera Localization in Sparse 3D Maps for GPS-Denied Vehicles
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing autonomous vehicle systems face challenges in accurately determining vehicle location in GPS-denied environments, such as parking garages, where traditional positioning systems like GPS and INS experience signal loss and drift, necessitating a reliable camera-based localization method for real-time navigation.
Innovation Solution
A camera-based localization technique that uses a 3D sparse map created through a teach and repeat scheme, where on-vehicle cameras capture images, detect features, and match them against pre-built map features to estimate the vehicle's pose, utilizing intrinsic and extrinsic calibration parameters from sensors like lidar and GPS to determine the vehicle's position and orientation in 3D space.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional positioning systems like GPS and INS are used for vehicle localization, then the system can operate in open environments, but the systems experience signal loss and drift in GPS-denied environments such as parking garages
Solution Approach 1:
The system segments the localization task into two parts: using GPS/INS for global positioning in open environments, and switching to camera-based visual localization in GPS-denied environments. This segmentation allows each subsystem to operate optimally in its suitable environment, resolving the contradiction between reliability and environmental adaptability.
Solution Approach 2:
The camera-based visual localization system acts as an intermediary when GPS signals are unavailable. The system uses cameras to capture images, extract feature points, and match them against pre-built 3D maps to determine vehicle position and orientation, providing continuous localization capability across different environmental conditions.
2Productivity
If camera-based localization is implemented for GPS-denied environments, then the system achieves real-time positioning capability, but the measurement precision and accuracy are challenged in sparse 3D mapped environments
Solution Approach 1:
The system performs preliminary actions by pre-building detailed 3D maps of the environment before the vehicle needs navigation. These maps contain pre-processed feature points and descriptors that can be quickly matched against live camera images, enabling real-time localization without sacrificing accuracy even in sparse environments.
Solution Approach 2:
The system changes parameters by adjusting the density and distribution of feature points in the 3D map based on environmental characteristics. In sparse environments, the system optimizes feature selection and matching parameters to maximize localization accuracy while maintaining real-time performance.
3Measurement precision
If a teach and repeat scheme with pre-built 3D maps is used, then the system achieves accurate localization, but the device complexity and computational requirements increase
Solution Approach 1:
The system creates simplified copies of the physical environment in the form of 3D maps containing feature points and descriptors. These digital copies are built once during a teaching phase and then reused repeatedly for localization, achieving high accuracy without requiring complex real-time processing of the full environmental model.
Solution Approach 2:
The computationally intensive tasks of 3D map building and feature extraction are performed in advance during a teach phase. During actual operation, the system only needs to perform lighter computational tasks like feature matching and pose estimation, reducing real-time computational requirements while maintaining high localization accuracy.
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.