Lidar Geometric Mesh Estimation for Noisy Drivable Surfaces
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing lidar systems struggle to accurately estimate drivable surfaces in an environment, as they often rely on incomplete or noisy point cloud data, leading to inaccuracies in determining drivable areas.
Innovation Solution
A lidar-based system that generates a geometric mesh from point cloud data, using azimuth and elevation measurements to triangulate a two-dimensional mesh, which is then expanded to three dimensions, and applies selection criteria to identify drivable surfaces by analyzing face normals and orientation, optimizing mesh updates based on environmental changes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional lidar systems use point cloud data for drivable surface estimation, then the system can operate with existing sensor data, but the estimation accuracy deteriorates due to noise and incomplete data
Solution Approach 1:
The patent introduces an intermediate geometric mesh structure that serves as a mediator between the raw point cloud data and the final drivable surface estimation. The mesh acts as a filtering layer that reconstructs continuous surfaces from discrete noisy points, thereby improving estimation accuracy while maintaining reliability
Solution Approach 2:
The system performs preliminary mesh generation and surface reconstruction before conducting drivable surface estimation. By pre-processing the point cloud data into a structured mesh format with inferred geometry, the system prepares cleaner, more reliable data for subsequent analysis, improving both accuracy and reliability
2Measurement precision
If the system generates and processes geometric mesh from point cloud data, then drivable surface estimation accuracy improves, but computational complexity increases
Solution Approach 1:
The patent segments the complex mesh processing task into distinct modules: point cloud to mesh conversion, mesh refinement, surface normal calculation, and drivable surface classification. This segmentation reduces overall complexity by making each sub-task more manageable and computationally efficient
Solution Approach 2:
The system applies partial mesh processing by focusing computational resources only on relevant portions of the mesh that contribute to drivable surface estimation, rather than processing the entire mesh uniformly. This selective approach reduces computational complexity while maintaining accuracy
3Measurement precision
If the system updates mesh based on environmental changes, then the estimation remains current and accurate, but processing time increases
Solution Approach 1:
The system implements periodic mesh updates triggered by environmental change detection rather than continuous updates. By monitoring for changes and updating only when necessary, the system maintains current and accurate estimations while minimizing processing time and computational overhead
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
The system provides accurate and efficient estimation of drivable surfaces by reducing noise and improving the precision of drivable area identification, allowing for optimized mesh updates based on environmental changes.
Implementation Method 1
The system determines the distance to the target based on the time of flight for a pulse of light emitted by the light source to travel to the target and back to the lidar system
Data Source
AI summary
A point cloud generated at least in part using a lidar sensor is received. A geometric mesh based on the point cloud is determined. A seed geometric face formed in the geometric mesh is selected based on one or more seed selection criteria. Starting from the seed geometric face, neighboring geometric faces of the geometric mesh that meet one or more relative neighbor selection criteria are iteratively selected into a region group, and an operable region indicated by the region group is detected.