Vision-Based Vehicle Localization for Urban Road Structure Matching
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing vehicle localization methods, such as GPS, face accuracy limitations in built-up urban areas due to signal degradation, and may not provide a reliable location on the map, especially for autonomous driving where precise road structure awareness is critical for decision-making.
Innovation Solution
The use of vision-based road structure detection using convolutional neural networks (CNNs) to identify and match road structures from images with pre-defined road maps, combining 3D imaging and particle filters to enhance localization accuracy, even in areas with occlusions or uncertainty.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If satellite positioning (GPS) is used for vehicle localization, then the location can be determined globally, but the accuracy degrades in built-up urban areas due to signal degradation
Solution Approach 1:
The patent combines satellite positioning data with vision-based road structure detection to create a hybrid localization system. The visual road structure detector captures images and identifies road features, which are then matched with pre-stored road map data to determine vehicle location independently of satellite signals, thereby compensating for GPS degradation in urban areas.
Solution Approach 2:
The patent introduces road structure matching as an intermediary method between direct satellite positioning and final location determination. When satellite signals are degraded, the system uses visual detection of road structures (lines, intersections, landmarks) as an intermediate step to infer vehicle location by matching these structures with the road map, providing a reliable fallback mechanism.
2Measurement precision
If vision-based road structure detection is used to improve localization accuracy, then the system complexity increases due to additional sensors and processing
Solution Approach 1:
The patent makes the image capture device serve multiple functions: it is used both for general autonomous navigation and specifically for road structure detection in localization. The same hardware infrastructure (cameras, processors) is leveraged for both driving decisions and location determination, avoiding the need for separate dedicated localization sensors and reducing overall system complexity.
Solution Approach 2:
The system uses the vehicle's own existing sensor suite (image capture devices already present for autonomous driving) to perform road structure detection for localization purposes. Rather than adding external specialized equipment, the system repurposes its inherent visual sensing capabilities to solve the localization problem, making the vehicle self-sufficient.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A vehicle localization system implements the following steps: receiving a predetermined road map; receiving at least one road image; processing the road image to identify therein road structure for matching with corresponding structure of the road map, and determine a vehicle location relative to the identified road structure; and using the determined vehicle location relative to the identified road structure to determine a location of the vehicle on the road map, by matching the road structure identified in the road image with the corresponding road structure of the predetermined road map, the identified road structure comprising: a centre line, wherein determining the vehicle location comprises determining a lateral separation between the vehicle and the centre line; and/or a junction region, wherein determining the vehicle location relative thereto comprises determining a longitudinal separation between the vehicle and the junction region in a direction along a road being travelled by the vehicle.