3D Environment Reconstruction Using Volumetric Voxel Processing

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing methods for generating 3D models from depth camera data face challenges due to low resolution and high noise in individual depth maps, leading to incomplete and computationally intensive surface reconstruction, which is not suitable for real-time processing in applications like immersive gaming and robotics.

Innovation Solution

A computer-implemented method using a 3D volume made of voxels, where each voxel is processed by a separate thread to determine its depth image location and update its value based on distance from the camera, allowing for parallelizable processing and efficient construction of a dense 3D model.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Manufacturing precision

If surface reconstruction algorithms (e.g., Poisson surface reconstruction) are used to combine multiple depth maps into a coherent 3D model, then the model accuracy and surface connectivity are improved, but the computational cost becomes too high to support real-time processing

Engineering Contradiction:
Improve3D model accuracyVSAvoidreal-time processing speed
Core Design Contradiction:
Manufacturing precisionVSProductivity

Solution Approach 1:

The patent segments the 3D reconstruction problem by representing the environment as a volumetric grid of voxels rather than processing point clouds globally. Each voxel can be independently processed and updated, allowing parallel computation across multiple threads while maintaining model accuracy through systematic volume integration of depth data.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent transitions from 2D depth map processing to 3D volumetric representation. By organizing depth information into a 3D voxel grid and processing along multiple planes (x, y, z dimensions), the system achieves both real-time performance through efficient memory access patterns and accurate surface reconstruction through volumetric integration.

Inventive Principle:
Principle #17Another dimension (Dimensionality change)

2Speed

If individual depth maps are used directly for 3D modeling, then the processing speed is fast, but the resolution is too low and noise is too high to generate accurate models

Engineering Contradiction:
Improveprocessing speedVSAvoidmodel resolution
Core Design Contradiction:
SpeedVSManufacturing precision

Solution Approach 1:

The patent merges multiple depth maps into a unified 3D volumetric model by integrating depth information across multiple frames into the voxel grid. This combination accumulates data from multiple sources, averaging out noise and filling gaps while maintaining real-time processing through efficient incremental updates rather than reprocessing all data.

Inventive Principle:
Principle #5Merging (Combining)

Solution Approach 2:

The patent performs preliminary organization of depth data into a voxel-based volume structure before surface extraction. By pre-processing depth maps into structured volumetric data with proper spatial indexing, the system enables both high-resolution reconstruction and fast subsequent processing operations.

Inventive Principle:
Principle #10Preliminary action

3Reliability

If point clouds from multiple frames are combined to infer surface geometry, then the completeness of the model is improved, but the computational intensity increases significantly

Engineering Contradiction:
Improvemodel completenessVSAvoidcomputational complexity
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

The patent uses a simplified voxel-based representation as a copy of the environment's geometric structure, avoiding the need to process complex point cloud data directly. Each voxel stores aggregated depth information, providing a complete yet computationally efficient representation that captures essential surface geometry without the complexity of individual point processing.

Inventive Principle:
Principle #26Copying

Data Source

PatentEP2671210B1Three-dimensional environment reconstruction
Publication Date: 2019.08.07 MICROSOFT TECHNOLOGY LICENSING LLC
  • EP2671210B1 patent drawingFigure 1
  • EP2671210B1 patent drawingFigure 2
  • EP2671210B1 patent drawingFigure 3

AI summary

Three-dimensional environment reconstruction is described. In an example, a 3D model of a real-world environment is generated in a 3D volume made up of voxels stored on a memory device. The model is built from data describing a camera location and orientation, and a depth image with pixels indicating a distance from the camera to a point in the environment. A separate execution thread is assigned to each voxel in a plane of the volume. Each thread uses the camera location and orientation to determine a corresponding depth image location for its associated voxel, determines a factor relating to distance between the associated voxel and the point in the environment at the corresponding location, and updates a stored value at the associated voxel using the factor. Each thread iterates through an equivalent voxel in the remaining planes of the volume, repeating the process to update the stored value.