Lane Recognition Device Using Local Voting for Noise Reduction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing lane recognition technologies face challenges in accurately extracting lane marks from images due to high calculation loads, noise sensitivity, and difficulty in distinguishing curved roads, especially when linear elements from non-lane marks are mixed with lane marks.
Innovation Solution
A lane recognition device that generates a linear element extraction image, sets local areas, calculates local straight lines, and uses a voting system to determine lane boundary lines by differentiating between lane marks and non-lane marks based on direction and intersection votes, with adjustable section numbers and threshold settings to optimize processing accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If Hough transform is used for lane mark detection, then lane marks can be detected, but calculation load becomes heavy
Solution Approach 1:
The image is divided into multiple local areas, and straight lines are detected independently in each local area rather than processing the entire image at once. This segmentation approach reduces the calculation load while maintaining detection accuracy for lane marks.
Solution Approach 2:
Different processing approaches are applied to different regions of the image. Local straight line detection is performed in each local area with parameters optimized for that specific region, improving overall detection efficiency and accuracy.
2Adaptability or versatility
If point information is used for detection, then detection flexibility is improved, but noise sensitivity increases
Solution Approach 1:
Multiple local straight line detection results are merged through voting to form the final lane mark detection. This combining approach reduces noise sensitivity by requiring consensus across multiple local areas, while maintaining detection flexibility through the voting mechanism.
3Quantity of substance
If conventional edge detection is used, then edge points can be extracted, but curved roads cannot be detected
Solution Approach 1:
The image is divided into multiple local areas along the road direction. By detecting straight lines in each local area and connecting them sequentially, the system can detect curved roads as a series of connected straight line segments, maintaining both edge extraction capability and curved road detection.
Solution Approach 2:
The detection method accommodates curved roads by allowing the orientation and parameters of straight lines to vary across different local areas. This enables the detection system to follow the curvature of the road while maintaining straight line detection in each local segment.
4Reliability
If detection straight lines from cutting-in vehicles are not filtered, then detection completeness is maintained, but lane boundary recognition accuracy decreases
Solution Approach 1:
A voting mechanism provides feedback to evaluate detection straight lines. Lines that do not receive sufficient votes from multiple local areas are identified as spurious detections (such as those from cutting-in vehicles) and are filtered out, improving lane boundary recognition accuracy while maintaining detection completeness.
Data Source
AI summary
Provided is a lane recognition device capable of extracting linear elements derived from lane marks from a linear element extraction image obtained by processing a captured image and recognizing lane boundary lines. Local areas 47 are set for a lane extraction area 45 set in a linear element extraction image 40 that each of linear elements is included in one or a plurality of the local areas 47, having a predetermined size, and a local straight line 44 of each local area is determined (vx) and (ψ) associated with the direction and the intersection x with a predetermined reference horizontal line 46 are calculated for each local straight line 44. Each local straight line 44 is defined as one vote, being casted to (vx, ψ) of a voting space 56. Lane boundary lines are recognized from detection straight lines, whose direction and the intersection x determined based on vote results.