Road Partition Detection Using Disparity Top View
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing road detection methods are inefficient in handling varying road heights and complex environments, particularly with sloping roads, as they often rely on single color images and fail to detect all types of road partitions flexibly, such as white lines and fences.
Innovation Solution
The method involves obtaining a disparity top view to detect parallel lines as road partitions, using points with zero disparity values as basic points and calculating included angles to determine road vanishing points, which allows for the detection of multiple road partitions simultaneously, regardless of height variations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional road detection methods using single color images or polarizing images are used, then the detection process is simple, but the detection accuracy deteriorates in cases of feeble road edge images or complicated environments with slanted roads
Solution Approach 1:
The patent combines multiple image data sources including color images, polarizing images, and disparity maps into a unified detection framework. By merging these different types of image information, the system achieves more accurate road partition detection in complex environments while maintaining a manageable system architecture through integrated processing.
Solution Approach 2:
The patent introduces disparity information from stereo vision as an additional dimension of data. By utilizing the disparity map that encodes depth information, the system can distinguish road partitions at different heights and slopes, significantly improving detection accuracy in three-dimensional road scenarios without requiring overly complex hardware.
2Adaptability or versatility
If modeling methods are applied to detect road shoulders, then detection is possible for those specific features, but it becomes inefficient to construct a special model for each road partition type such as white lines or fences
Solution Approach 1:
The patent employs a universal detection algorithm based on parallel line detection in disparity top views that can handle multiple types of road partitions including road shoulders, white lines, and fences. This single method replaces the need for multiple specialized models, achieving versatile detection across different partition types while simplifying the overall system architecture.
3Measurement precision
If three-dimensional coordinates are projected as two-dimensional coordinates along a direction across the road to detect lane lines, then lane line detection is achieved, but the method becomes invalid in cases where height of the road is different, like a sloping road
Solution Approach 1:
The patent utilizes the disparity dimension from stereo vision to create a disparity top view that preserves height information. By detecting parallel lines in this disparity space rather than projecting to a simple 2D view, the method maintains accuracy for road partitions at different heights and slopes, effectively solving the limitation of traditional projection methods.
Solution Approach 2:
The patent changes the detection parameter space by working in disparity coordinates rather than simple image coordinates. By transforming the detection problem into the disparity domain where height variations are encoded as distinct patterns, the method achieves consistent detection accuracy across varying road heights and slopes without requiring complex adaptive models.
Data Source
AI summary
A method and an apparatus are used for detecting a road partition, the method comprising: a step of obtaining a disparity top view having a road area; and a step of detecting parallel lines as the road partitions from the disparity top view.