Spherical Range Map Generation for Efficient Autonomous Location Recognition
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for generating maps for autonomous driving face challenges such as inaccurate environmental modeling, high data storage requirements, sensitivity to initial values, reduced precision in certain conditions, and increased computational complexity, particularly with 3D point clouds and road surface pattern methods.
Innovation Solution
A method and apparatus that generate a three-dimensional map by projecting 3D coordinate information onto a 2D plane to create a spherical range image, perform semantic segmentation, and synthesize lane attribute information, using a graph-based structure to optimize data storage and processing.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If 3D point cloud is used as map data, then location recognition precision is improved, but data storage space requirement increases significantly
Solution Approach 1:
The patent creates a simplified 2D projection copy of the 3D point cloud environment onto a spherical surface. This spherical range image preserves the essential spatial relationships and features needed for location recognition while occupying significantly less storage space than the original 3D point cloud data.
Solution Approach 2:
The patent transforms 3D spatial data into a 2D spherical projection representation. By mapping 3D coordinate information onto a spherical surface and then projecting it to 2D, the system reduces dimensional complexity while maintaining the geometric relationships necessary for accurate location recognition through gradient calculations.
2Measurement precision
If 3D point cloud is used as map data, then location recognition precision is improved, but computational complexity increases
Solution Approach 1:
The patent reduces computational complexity by transforming 3D point cloud processing into 2D spherical projection processing. Gradient calculations are performed on the simplified spherical range image rather than the complex 3D point cloud, significantly reducing the computational burden while preserving location recognition accuracy.
Solution Approach 2:
The patent extracts only the essential features needed for location recognition from the full 3D point cloud data. By creating a spherical range image that captures the critical spatial relationships and gradients, the system eliminates unnecessary computational overhead associated with processing complete 3D point cloud information.
3Measurement precision
If HD map with lane-level precision is used, then location recognition is improved, but environmental modeling accuracy deteriorates
Solution Approach 1:
The patent applies different levels of detail to different aspects of the map representation. The spherical range image provides accurate local geometric information for location recognition through gradient calculations, while the overall environmental modeling maintains sufficient fidelity for autonomous navigation without requiring exhaustive lane-level detail everywhere.
Data Source
AI summary
Disclosed are a method and an apparatus for generating a map for autonomous driving and recognizing a location based on the generated map. When generating a map, a spherical range image is obtained by projecting 3D coordinate information corresponding to a 3D space onto a 2D plane, and semantic segmentation is performed on the spherical range image to generate a semantic segmented image. Then, map data including a spherical range image, a semantic segmented image, and lane attribute information are generated.