Image-Based Lane Indexing for GNSS-Limited Vehicle Localization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing autonomous vehicle localization methods rely on Global Navigation Satellite System (GNSS), inertial measurement units, and digital maps, which can be computationally expensive, unreliable in areas with poor signal reception, or prone to significant errors.
Innovation Solution
The use of machine learning models that generate lane indices based on real-time image data from sensors, such as cameras and LIDAR, to improve lane offset detection and vehicle localization, reducing reliance on potentially inaccurate location data sources.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If existing localization methods using GNSS, inertial measurement units, and digital maps are employed, then comprehensive positioning capability is achieved, but computational cost increases and reliability decreases in areas with poor signal reception
Solution Approach 1:
The patent extracts and removes the reliance on GNSS and complex digital map matching systems, replacing them with a simplified vision-based lane offset detection system that processes only image data from onboard cameras to determine vehicle position
Solution Approach 2:
The patent substitutes the mechanical and computational systems (GNSS receivers, inertial measurement units, digital map databases) with an optical system using cameras and machine learning algorithms to achieve localization through image processing
2Measurement precision
If existing localization methods using GNSS and digital maps are used, then positioning coverage is maintained, but measurement precision deteriorates in areas with reduced signal reception
Solution Approach 1:
The patent introduces lane detection algorithms and machine learning models as intermediary systems that process visual information from lane markings and road features to infer vehicle position, serving as a mediator between the camera sensors and the localization output
Solution Approach 2:
The patent changes the fundamental parameter for localization from satellite signal strength and digital map coordinates to visual features extracted from road imagery, such as lane marking patterns, road geometry, and perspective transformations
3Measurement precision
If machine learning models using image data are implemented, then localization accuracy in poor GNSS areas improves, but device complexity increases
Solution Approach 1:
The patent segments the localization task into distinct processing stages: image capture from cameras, preprocessing to extract road features, machine learning model inference to detect lane markings and estimate offset, and post-processing to convert pixel coordinates to real-world measurements
Solution Approach 2:
The patent creates a simplified digital representation of the road scene by copying and processing only the essential visual features (lane markings, road edges) rather than processing complete high-resolution images, reducing computational complexity while maintaining accuracy
Data Source
AI summary
Systems and methods for training and executing machine learning models to generate lane index values are disclosed. A method includes identifying a set of image data captured by at least one autonomous vehicle when the at least autonomous vehicle is positioned in a lane of a roadway and respective ground truth localization data; determining a plurality of lane index values for the set of image data based on the ground truth localization data; labeling the set of image data with the plurality of lane index values, the lane index values representing a number of lanes from a leftmost or rightmost lane to the lane in which the at least one autonomous vehicle was positioned; and training, using the labeled set of image data, a plurality of machine learning models that generate a left lane index value and a right lane index value as output.