Virtual Lane Mark Generation for Worn or Obscured Road Markings
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous vehicles face difficulties in navigating through road segments with unclear or worn lane marks, especially in adverse environmental conditions like rain or snow, as existing methods struggle to accurately discern lane markings from camera images.
Innovation Solution
A system and method that generate virtual lane marks by determining the road edge and lane attributes using camera images, transforming them into a bird's eye view, and employing neural networks (such as multilayer perception, autoencoder/decoder, or convolutional neural networks) to create and track virtual lane marks, which can include information from GPS, map servers, and surrounding agents' trajectories.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional camera image analysis is used to identify lane marks, then the system is simple and easy to implement, but it fails to accurately discern lane marks in unfavorable conditions such as rain, snow, or worn markings
Solution Approach 1:
The patent introduces virtual lane marks as an intermediary element that mediates between the physical road infrastructure and the autonomous vehicle's navigation system. These virtual marks, generated through neural network processing of camera images and map data, serve as a reliable reference for lane tracking when physical marks are indistinguishable due to weather conditions, wear, or environmental factors.
Solution Approach 2:
The patent creates a virtual copy of lane marks through neural network-generated images that replicate the appearance and position of actual lane markings. This copying approach allows the system to work with idealized lane mark representations even when the physical marks are degraded or invisible, thereby maintaining navigation reliability without requiring complex hardware modifications.
2Measurement precision
If virtual lane marks are generated using neural networks and multiple data sources, then navigation accuracy improves in unfavorable conditions, but the computational complexity and processing requirements increase
Solution Approach 1:
The patent performs preliminary processing of camera images through neural networks to generate virtual lane marks before they are needed for navigation decisions. By pre-computing the lane mark positions and appearances under various conditions, the system prepares accurate reference data in advance, reducing the computational burden during real-time navigation and allowing complex processing to occur before critical decision points.
Data Source
AI summary
A vehicle, system method for operating the vehicle is disclosed. The system includes a camera and a processor. The camera is configured to obtain a camera image of a road segment. The processor determines a location of a road edge for the road segment within the camera image, obtains a lane attribute for the road segment, generates a virtual lane mark for the road segment based on the road edge and the lane attribute, and moves the vehicle along the road segment by tracking the virtual lane mark.