Movable Area Boundary Fusion for Reliable Environment Mapping
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing techniques for detecting the movable area of a mobile body, such as vehicles or robots, are prone to errors due to shadows, noise, and false points from laser radar, leading to inaccurate occupancy grid maps.
Innovation Solution
A signal processing device and method that incorporates a labeling boundary estimation unit, a distance image boundary estimation unit, and a movable area boundary determination unit to accurately detect the movable area by combining labeling information from images and distance information from depth sensors, generating a precise environment map for autonomous movement.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If edge detection is performed based on image captured by camera, then road surface boundary can be extracted, but detection accuracy deteriorates due to shadows, noise, and false points from laser radar
Solution Approach 1:
The patent combines labeling information from image processing with distance information from laser radar to determine the movable area boundary. By merging these two independent information sources, the system achieves more reliable boundary detection that overcomes the limitations of using either method alone, particularly in challenging conditions with shadows and noise.
2Measurement precision
If occupancy grid map is generated using laser radar point cloud information, then three-dimensional structure can be determined, but measurement accuracy deteriorates due to false points and noise
Solution Approach 1:
The patent uses labeling information from image processing as an intermediary to filter and validate distance information from laser radar. The labeling information acts as a reference that helps identify and eliminate false points in the point cloud data, thereby improving the reliability of the occupancy grid map generation.
3Measurement precision
If multiple information sources are combined for boundary detection, then detection accuracy improves, but device complexity increases
Solution Approach 1:
The patent creates a unified boundary determination mechanism that handles multiple information sources (image labeling and distance data) through a single integrated process. This multi-functional approach allows the system to process different types of data using common algorithms, reducing the overall complexity compared to maintaining separate processing chains for each data source.
Data Source
AI summary
On the basis of semantic labeling information in an image captured by a camera an area in which a boundary of a movable area of a mobile body exists in the image is estimated as an image movable area boundary, on the basis of a distance image detected by a depth sensor an area in which the boundary of the movable area of the mobile body exists in the distance image is estimated as a distance image movable area boundary, the boundary of the movable area is determined on the basis of the image movable area boundary and of the distance image movable area boundary, and an environment map is generated on the basis of the determined boundary of the movable area. The present disclosure can be applied to an in-vehicle system.