Lane Line Fusion Using Map Data for Missing and False Detections
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing lane line detection methods in autonomous driving face challenges in complex scenarios, leading to false and missing detections due to unclear or blocked lane lines, with poor generalization capabilities and unstable performance across different lighting and ground conditions.
Innovation Solution
A method that merges lane line information from a map server with detected information to improve accuracy, using a first apparatus to determine third lane line information based on first and second lane line information, and optionally incorporating historical data to enhance detection performance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If deep learning method is used to improve generalization performance, then detection accuracy is improved, but false detection and missing detection still occur in complex scenarios
Solution Approach 1:
The patent merges map information (prior knowledge) with real-time detection information (sensor data) to form fused lane line information. This combination allows the system to leverage the generalization capability of deep learning while using map data to correct false detections and fill missing detections in complex scenarios, thereby improving both accuracy and reliability simultaneously
Solution Approach 2:
Map information serves as an intermediary between prior knowledge and real-time sensor data. The map data acts as a reference framework that mediates the detection process, helping to resolve ambiguities in complex scenarios where direct detection fails, thus improving detection stability without sacrificing accuracy
2Measurement precision
If artificially set features are used for lane segmentation, then performance is good in highway scenarios, but generalization capability is very poor
Solution Approach 1:
The patent creates a universal lane line detection system that can adapt to multiple scenarios (highway, urban, rural, different lighting conditions) by combining map information with real-time detection. The system performs the function of both using artificial features for structured environments and deep learning for generalization, making it versatile across different driving scenarios
Solution Approach 2:
Map information is prepared in advance as prior knowledge before real-time detection occurs. This preliminary action of having preprocessed map data available allows the system to quickly reference and compare against actual sensor data, improving both accuracy in specific scenarios and generalization to new scenarios
3Speed
If detection is performed in complex scenarios with blocked or unclear lane lines, then real-time detection capability is maintained, but large quantity of false detection and missing detection occurs
Solution Approach 1:
The system uses map information as feedback to correct real-time detection results. When detection algorithms produce false or missing detections in complex scenarios, the map data provides feedback information about expected lane line positions and characteristics, allowing the system to correct errors while maintaining real-time performance
Solution Approach 2:
Map information serves as a cushioning mechanism that prepares the system in advance for potential detection failures. By having map data ready as a safety net, the system can compensate for false and missing detections that occur in complex scenarios, maintaining both real-time capability and accuracy
Data Source
Figure 1~3
Figure 4
Figure 5
AI summary
A lane line information determining method and apparatus are disclosed, and are applicable to the field of autonomous driving or intelligent driving. The method includes: A first apparatus obtains first lane line information corresponding to a location of a vehicle, where the first lane line information is from a map server (201). The first apparatus obtains second lane line information (202). The first apparatus determines third lane line information based on the first lane line information and the second lane line information (203). Lane line information obtained from the map server is merged with detected lane line information, so that false detection is eliminated and information about a lane line whose detection is missing is retrieved, to help improve lane line detection performance, accurately determine lane line information, and improve driving performance and safety.