Vehicle Road Surface Recognition via Segmented 3D Modeling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing vehicle exterior environment recognition systems face challenges in accurately identifying road surfaces, particularly those distant from the vehicle, due to low reliability of images and parallax, which can lead to improper determination of road gradients or failure in extracting three-dimensional objects.
Innovation Solution
The system employs a vehicle exterior environment recognition apparatus with a first and second road surface identifier to generate road surface models by plotting representative distances of horizontal arrays of blocks, using pattern matching and stereo methods to derive parallax information, and applies the Hough transform to filter noise, ensuring accurate identification of road surfaces and three-dimensional objects.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If the system uses image-based road surface identification, then the road surface can be detected, but the reliability is low for distant road surfaces
Solution Approach 1:
The patent divides the road surface detection task into multiple segments by creating multiple road surface models from different horizontal arrays of blocks at various vertical positions. Each model represents a specific depth range, allowing the system to systematically address distant road surface detection by processing it in manageable segments rather than as a single difficult task.
Solution Approach 2:
The patent transitions from two-dimensional image analysis to three-dimensional road surface modeling by plotting representative distances at respective vertical positions. This dimensional transformation enables the system to represent distant road surfaces in a 3D space, improving detection reliability by adding depth information and creating a more comprehensive spatial model.
2Measurement precision
If the system plots representative distances of horizontal arrays of blocks, then road surface models are generated, but noise filtering is required
Solution Approach 1:
The patent introduces the Hough transform as an intermediary filtering mechanism between data collection and final road surface model generation. This intermediary process transforms the raw distance measurements into a standardized representation, automatically filtering noise while preserving accurate measurements, thus improving precision without proportionally increasing system complexity.
3Loss of information
If the system uses stereo methods to derive parallax information, then depth information is obtained, but image reliability is reduced for distant objects
Solution Approach 1:
The patent applies preliminary filtering and processing to the parallax information before final road surface model generation. By pre-processing the distance measurements and applying the Hough transform to filter noise early in the pipeline, the system preserves accurate parallax information while compensating for the reduced reliability of distant images through systematic error reduction.
Data Source
AI summary
A vehicle exterior environment recognition apparatus includes a first road surface identifier and a second road surface identifier. The first road surface identifier identifies a first road surface region of a road surface in an image and generates a first road surface model by plotting, at respective vertical positions, representative distances of horizontal arrays of blocks in the first road surface region. The second road surface identifier identifies a second road surface region in the image and generates a second road surface model by plotting, at respective vertical positions, representative distances of horizontal arrays of blocks in the second road surface region. The second road surface region is a region of the road surface that is farther from a vehicle than the first road surface region is, and has a horizontal length greater than a horizontal length of the first road surface region in a three-dimensional space.