3D Step Position Detection on Sloped and Occluded Road Surfaces
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods struggle to accurately detect low positions on a road surface using parallax images from a stereo camera, especially when the shape of the low position changes or there is a large slope angle, leading to difficulties in detecting steps due to occlusion and varying distances.
Innovation Solution
A step position detection apparatus that utilizes a point group information acquisition unit, such as a stereo camera, to acquire road surface data, estimate flat roads, and detect steps by creating a two-dimensional map divided by grid spaces, calculating road surface slopes, and extending hidden road surfaces to accurately determine step positions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If parallax images from a stereo camera are used to detect low positions on the road surface, then three-dimensional information can be obtained, but detection accuracy deteriorates when the shape of the low position changes or when there is a large slope angle
Solution Approach 1:
The patent transitions from two-dimensional parallax image data to three-dimensional point group information by introducing height data. This dimensional enhancement allows the system to accurately represent road surface topology including steps and slopes, resolving the contradiction between maintaining detection accuracy and adapting to varying shapes.
Solution Approach 2:
The patent changes the data representation parameters from simple parallax displacement to comprehensive point group information including (x, y, z) coordinates and height values. This parameter transformation enables accurate detection of low positions regardless of shape changes or slope angles, simultaneously improving both measurement precision and adaptability.
2Device complexity
If conventional parallax calculation methods are used, then processing is simpler, but detection reliability deteriorates due to occlusion and varying distances
Solution Approach 1:
The patent performs preliminary classification of grid spaces into flat road areas and step areas before detailed analysis. This preliminary action organizes the complex three-dimensional data structure, making subsequent detection more reliable while managing processing complexity through systematic categorization.
Solution Approach 2:
The patent divides the road surface into discrete grid spaces and further segments them into different types (flat road, step area) based on height variations. This segmentation approach breaks down the complex detection problem into manageable units, improving reliability by analyzing each segment independently while controlling overall processing complexity.
3Measurement precision
If detailed three-dimensional analysis is performed to accurately detect steps, then detection precision improves, but processing time increases
Solution Approach 1:
The patent segments the road surface into grid spaces and classifies them into flat road areas and step areas. This segmentation allows the system to apply different processing strategies to different regions, achieving precise step detection in step areas while avoiding unnecessary processing in flat areas, thus reducing overall processing time.
Solution Approach 2:
The patent applies different analysis methods to different regions: simple height threshold checking for flat road areas and extended surface analysis for step areas. This local quality approach ensures high detection precision where needed while minimizing processing time in regions requiring less analysis.
Data Source
AI summary
A step position detection apparatus includes a processor; and a memory that includes instructions, which when executed, cause the processor to execute: acquiring, by an imaging device, a condition including at least a road surface around a work machine, as point group information; estimating an area where unevenness of the road surface does not exceed a predetermined height, as a flat road based on the acquired point group information including height information; and estimating an extended part of the road surface hidden by a step, among a first flat road where the work machine exists and a second flat road located at a lower position than the first flat road included in the estimated flat road.