Lane Detection Using Multi-ROI Resolution for Remote Road Accuracy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current lane detection technologies face challenges in accurately detecting lanes from remote road regions due to reduced image resolution for processing speed improvements, leading to incomplete lane recognition, especially in autonomous driving systems.
Innovation Solution
The proposed lane detection method sets multiple regions of interest (ROIs) in a forward-view image, allowing for accurate lane pixel detection using a neural network-based detector, where the image resolution of the basic ROI is reduced for faster processing, while maintaining original resolution for remote road regions to enhance detection accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If image resolution is reduced for processing speed improvement, then processing speed is improved, but lane detection accuracy deteriorates
Solution Approach 1:
The patent divides the forward-view image into multiple regions of interest (ROIs) based on distance from the vehicle. Close road regions use reduced resolution for speed, while remote road regions maintain original resolution for accuracy. This segmentation allows simultaneous optimization of processing speed and detection accuracy in different spatial zones.
Solution Approach 2:
Different resolution qualities are applied to different regions of the image. Remote road regions, which are more critical for safe autonomous driving, maintain original high resolution. Close road regions use reduced resolution. This local quality differentiation ensures that computationally intensive areas receive appropriate processing power while maintaining overall system efficiency.
2Measurement precision
If multiple regions of interest are set with different resolutions, then lane detection accuracy is improved, but device complexity increases
Solution Approach 1:
The image processing system is segmented into multiple ROI zones with different resolution requirements. This segmentation is implemented through software algorithms that identify and process different regions separately, avoiding the need for complex hardware modifications while achieving superior detection accuracy.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A lane detection method and apparatus is disclosed. The lane detection apparatus may set a regions of interest (ROIs) in a forward-view image of a vehicle obtained by a camera, estimate lane pixels corresponding to a lane from each of the ROIs using a neural network-based detector, and determine a lane region in the forward-view image based on the lane pixels.