Road Lane Geometry Estimation Using Camera and Leading Vehicle Traces
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for estimating road lane geometry in semi-autonomous and autonomous driving systems are limited by the short range of camera detection and the inability of radar to detect individual lane markings, making it difficult to accurately infer road lane geometry over extended distances.
Innovation Solution
A method and system that combines camera-based lane estimation using visible and near-infrared cameras with leading vehicle tracking, utilizing artificial intelligence to stitch camera-estimated and leading-vehicle-estimated lane segments, extending the range of road lane geometry estimation up to 300 meters and improving accuracy by smoothing and extrapolating data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If camera detection is used to estimate road lane geometry, then measurement precision is improved, but the detection range is limited to at most 100 m
Solution Approach 1:
The patent combines camera-based lane estimation with leading vehicle-based lane estimation to create a hybrid system. The camera provides high-precision lane geometry data within its limited range (up to 100m), while the leading vehicle tracking extends the detection range up to 300m by inferring lane geometry from vehicle traces, effectively merging the strengths of both methods to overcome the range limitation of camera-only systems.
2Length of stationary object
If radar is used to detect road geometry, then the detection range is extended, but the ability to detect individual lane markings is lost
Solution Approach 1:
The patent uses leading vehicles as an intermediary to bridge the gap between radar's long-range detection capability and the need for precise lane marking information. By tracking the positions and traces of leading vehicles, the system infers lane geometry at distances beyond camera range while maintaining lane-level precision, effectively using vehicles as mediators to transfer lane information over extended ranges.
3Measurement precision
If only camera-based lane estimation is used, then lane geometry accuracy is maintained, but the range is insufficient for semi-autonomous and autonomous driving requirements
Solution Approach 1:
The system performs preliminary lane geometry estimation using cameras at short ranges (0-100m) where high precision is available, while simultaneously using leading vehicle traces to establish lane geometry at extended ranges (100-300m). This preliminary action at multiple ranges allows the system to maintain reliability for autonomous driving by having accurate lane information available throughout the entire operating range, not just within camera limits.
Data Source
AI summary
A method and system for estimating road lane geometry includes a camera-estimated lane segment, for estimating lane geometry based on camera detection of road markings and a leading-vehicle-estimated lane segment, for estimating lane geometry based on traces of at least one leading vehicle. Estimated road geometry is obtained from a combination of the camera-estimated lane segment and the leading-vehicle-estimated lane segment.