Road Lane Geometry Estimation by Stitching Camera and Vehicle Traces
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for estimating road lane geometry are limited in range, with cameras covering up to 100 m and radar unable to detect individual lanes, posing challenges for semi-autonomous and autonomous driving systems.
Innovation Solution
A method combining camera-based lane estimation from road markings and leading vehicle tracking to extend the range of lane detection, using visible and near-infrared cameras and artificial intelligence to identify road markings and vehicle traces, with stitching and smoothing techniques to enhance accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If camera-based lane detection is used, then lane marking detection accuracy is improved, but detection range is limited to 100 m
Solution Approach 1:
The patent combines camera-based lane detection (providing accurate lane marking detection up to 100m) with radar-based leading vehicle detection (providing detection up to 300m) to create a hybrid system that achieves both accurate lane identification and extended detection range. The camera detects road markings while the radar tracks leading vehicles, and their results are merged to provide comprehensive lane geometry information.
2Length of moving object
If radar is used for detection, then detection range is extended to entire road ahead, but ability to detect individual lanes is lost
Solution Approach 1:
The patent uses leading vehicles as intermediaries to bridge the gap between radar's long-range detection capability and the need for individual lane identification. The radar detects leading vehicles at long ranges, and their positions and trajectories serve as proxies for lane geometry, allowing the system to infer individual lane information from radar data that would otherwise only provide overall road geometry.
3Measurement precision
If only camera detection is used, then lane marking accuracy is maintained, but range coverage is insufficient for autonomous driving requirements
Solution Approach 1:
The patent creates a multi-functional detection system where the camera serves for accurate lane marking detection in the near field, while the radar serves for long-range vehicle detection and lane inference. The system universally handles both short-range precise lane geometry and long-range road structure detection, making it adequate for autonomous driving requirements that span multiple distance scales.
Data Source
Figure 1~2
Figure 3a
Figure 3b
AI summary
The invention relates to a method for estimating road lane geometry (14). A camera lane estimation comprising at least one camera-estimated lane segment (6), based on camera (3) detection of road markings is provided. Further, a leading vehicle lane estimation comprising at least one leading-vehicle-estimated lane segment (11), based on traces (10) of at least one leading vehicle (9) is provided. Then, the at least one camera-estimated lane segment (6) and the at least one leading-vehicle-estimated lane segment (11) are stitched together to obtain the estimated road lane geometry (14). Further, a system (2) for estimating road lane geometry (14) and a vehicle (1) comprising such a system (2) are provided.