Lane Line Anomaly Recognition Using Geometric Verification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing lane line recognition technologies in autonomous vehicles often misidentify road signs as lane lines due to the use of deep learning schemes based on neural networks, leading to low accuracy in lane line detection.
Innovation Solution
A method for recognizing anomalous lane lines using convolutional neural networks (CNN) to process lane line images, involving length thresholding, marking straight lines, and determining lane spacing areas to filter out non-lane lines, thereby improving detection accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Extent of automation
If deep learning schemes based on neural networks are used for lane line detection, then the system can process image information automatically, but road signs are easily misidentified as lane lines, resulting in low accuracy
Solution Approach 1:
The patent segments the lane line detection process into multiple stages: initial detection by neural network, followed by geometric verification including length threshold checking, straightness verification, and lane spacing validation. This multi-stage segmentation allows automatic processing while improving accuracy by filtering out false positives like road signs through geometric constraints
Solution Approach 2:
The patent introduces geometric verification mechanisms as intermediary steps between the neural network detection and final lane line recognition. These intermediaries (length thresholds, straightness checks, lane spacing calculations) act as filters that verify the plausibility of detected lines, preventing road signs from being misidentified as lane lines while maintaining automated processing
2Measurement precision
If geometric verification methods are added to filter out false lane lines, then detection accuracy is improved, but the processing complexity and time increase
Solution Approach 1:
The patent applies preliminary geometric constraints (length thresholds, straightness requirements) to quickly eliminate obviously invalid detections before performing more complex verification. This preliminary filtering reduces the computational burden of subsequent processing steps while maintaining high accuracy by eliminating false positives early in the pipeline
3Measurement precision
If multiple verification steps are performed to eliminate false lane lines, then recognition accuracy is enhanced, but the processing time increases
Solution Approach 1:
The patent implements a hierarchical verification process where not all verification steps are applied equally to all detected lines. Simple geometric checks (length, straightness) are applied to all detections, while more complex lane spacing verification is applied selectively. This partial application of verification steps maintains high accuracy while reducing overall processing time compared to applying all verifications uniformly
Data Source
AI summary
The present disclosure relates to a field of intelligent driving technology field, and provides a method for recognizing anomalous lane line and an electronic device obtaining lane line recognition results of lane line images. The method includes: determining lane lines in the lane line images according to the lane line recognition results; determining one of the lane lines having a length greater than a first length threshold and having a straight line from the lane lines as the first lane line; masking the first lane line and determining mark points of the first lane; determining marking lines according to the mark points and the first lane line; dividing each of the lane line images according to the marking lines and obtaining a lane spacing area; and determining the lane lines in the lane spacing area of the lane line recognition results as abnormal lane lines.