Road Element Association Inference Using Prior Map Features
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current high-definition maps for automated driving are complex and costly to produce, have limited coverage, and do not update in real-time, leading to potential errors and restricted application scope due to reliance on outdated positioning information, with limited focus on parsing association relationships between road elements.
Innovation Solution
A method for determining association relationships between road elements using road perception and prior element information, processed to obtain feature vectors and verified through matching cost functions, allowing for robust and accurate inference of current road element relationships without strong reliance on outdated prior information.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If high-definition map production is used to obtain association relationships between road elements, then the association relationships can be obtained, but the production process is complex and costly with limited coverage and slow update speed
Solution Approach 1:
The patent uses aerial images as a copy or alternative representation of road scenes to extract association relationships between road elements, replacing the need for complex high-definition map production. The image processing system directly captures and analyzes road element relationships from aerial imagery, achieving the same informational goal through a simpler, more scalable approach.
2Reliability
If high-definition map is used for automated driving, then road structure information is available, but highly accurate positioning information is required which may cause errors when deviation occurs
Solution Approach 1:
The patent introduces aerial images as an intermediary layer between the vehicle and road structure information. Instead of directly relying on positioning accuracy to query pre-stored high-definition maps, the system uses image-based road element detection that is less sensitive to positioning deviations, providing a more robust alternative pathway to obtain road structure information.
3Productivity
If road structure cognition focuses only on geometric information reconstruction, then geometric data is obtained, but association relationships between road elements are not parsed
Solution Approach 1:
The patent merges geometric information reconstruction with association relationship parsing into a unified image processing framework. By simultaneously detecting road elements (lanes, intersections, traffic lights) and their spatial relationships from aerial images, the system achieves both geometric reconstruction and relationship extraction in one process, eliminating the need for separate analysis steps.
Data Source
Figure 1
Figure 2a~2b
Figure 2c
AI summary
Embodiments of this application provide a method for determining an association relationship between road elements, a related system, and a storage medium. The method may include: obtaining road perception element information and road prior element information; processing the road perception element information and the road prior element information to obtain a first feature vector and a second feature vector; and obtaining an association relationship between road perception elements and/or an association relationship between the road perception element and a road prior element based on the first feature vector and the second feature vector. In this manner, a current road element is associated with road prior information, so that the road prior information can serve as a bridge to improve a current road element relationship inference level. In addition, a redundant relationship can be constructed through inference of a relationship between the current road element and prior road information, to facilitate verification of a current road element relationship, thereby enhancing robustness and interpretability of an overall solution.