Poisson Surface Reconstruction for Sparse 3D Ground Meshes

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing 3D scanning technologies, particularly those using LIDAR data, struggle with generating detailed and textured mesh models of ground surfaces due to low point density and inability to handle occlusions, leading to inaccurate environmental maps for autonomous vehicles.

Innovation Solution

A system utilizing a Gaussian process to filter out outliers, add additional points to the point cloud, and perform Poisson surface reconstruction to generate an implicit surface, followed by polygon mesh representation, including methods like marching cubes and Bloomenthal Polygonizer, to create a detailed mesh model.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Manufacturing precision

If mesh models are generated from LIDAR point cloud data, then a surface representation is obtained, but the models lack detail and texture due to low point density

Engineering Contradiction:
Improvemesh model detailVSAvoidpoint cloud density
Core Design Contradiction:
Manufacturing precisionVSQuantity of substance

Solution Approach 1:

The patent applies preliminary action by performing outlier removal and point cloud completion before mesh generation. The Gaussian process-based outlier removal eliminates noisy points that degrade quality, while the point cloud completion algorithm adds virtual points in occluded regions, preparing enhanced input data for subsequent mesh reconstruction to achieve detailed models despite sparse LIDAR coverage

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The patent introduces an intermediary processing stage between raw LIDAR data and final mesh generation. This intermediary includes Gaussian process regression to model the ground surface probability distribution and synthesize missing points, acting as a mediator that transforms sparse point cloud data into a denser representation suitable for high-quality mesh reconstruction

Inventive Principle:
Principle #24Intermediary (Mediator)

2Reliability

If mesh models are generated from LIDAR data, then a surface representation is obtained, but the models cannot handle ground surfaces with occlusions

Engineering Contradiction:
Improveocclusion handling capabilityVSAvoidmissing LIDAR data
Core Design Contradiction:
ReliabilityVSLoss of information

Solution Approach 1:

The patent applies copying by generating virtual point cloud data that replicates the appearance and geometry of occluded ground surfaces. The Gaussian process model copies the statistical properties of visible ground surfaces to synthesize missing points in occluded regions, creating a complete point cloud representation that enables reliable mesh generation even when direct LIDAR measurements are unavailable

Inventive Principle:
Principle #26Copying

Solution Approach 2:

The patent performs preliminary completion of the point cloud data before mesh generation by using Gaussian process regression to predict and fill missing points in occluded regions. This preliminary action ensures that the input data for mesh reconstruction is complete and reliable, preventing failures in handling occluded ground surfaces

Inventive Principle:
Principle #10Preliminary action

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 produces a detailed and textured mesh model of ground surfaces, enhancing the accuracy of environmental maps for autonomous vehicles, enabling safer navigation and control.

Implementation Method 1

identify and discard one or more outliers in the 3D point cloud to generate a filtered point cloud using a Gaussian process

Methodology Applied
Scientific EffectGaussian process:

Implementation Method 2

use Poisson surface reconstruction to generate an implicit surface corresponding to the surface from the reconstruction dataset

Methodology Applied
Scientific EffectPoisson surface reconstruction:

Data Source

PatentEP4078535B1Methods and systems for constructing map data using poisson surface reconstruction
Publication Date: 2025.08.13 VOLKSWAGEN GROUP OF AMERICA INVESTMENTS LLC
  • EP4078535B1 patent drawingFigure 1
  • EP4078535B1 patent drawingFigure 2
  • EP4078535B1 patent drawingFigure 3

AI summary

A method and a system for generating a mesh representation of a surface. The method includes receiving a three-dimensional (3D) point cloud representing the surface, identifying and discarding one or more outliers in the 3D point cloud to generate a filtered point cloud using a Gaussian process, adding one or more additional points to the filtered point cloud to generate a reconstruction dataset, and using Poisson surface reconstruction to generate an implicit surface corresponding to the surface from the reconstruction dataset.