3D Surface Reconstruction Using Volume-Based Grid Processing

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current methods for real-time 3D surface reconstruction using multi-camera systems face challenges such as high computational complexity, limited resolution flexibility, and susceptibility to calibration errors, especially when trying to achieve high spatial resolutions.

Innovation Solution

A method that combines volume-based and surface-based approaches by discretizing the workspace, indexing 3D grid points, and using a mapping model to determine 3D point stacks, which are then processed to reconstruct the object surface with reduced computational effort and increased robustness, allowing for real-time reconstructions at higher resolutions without requiring complex hardware or extensive parallelization.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If volume-based methods are used for 3D surface reconstruction, then robustness against calibration errors is improved, but computational complexity increases cubically with spatial resolution

Engineering Contradiction:
Improverobustness against calibration errorsVSAvoidcomputational complexity
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

The workspace is discretized into a 3D grid of volume elements (voxels), dividing the continuous space into discrete manageable units. This segmentation allows the reconstruction algorithm to process only relevant voxels near the object surface rather than the entire workspace, reducing computational complexity from cubic to quadratic while maintaining volume-based robustness.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent extracts and processes only the near-surface volume elements that are relevant for surface reconstruction, discarding the interior voxels that do not contribute to the surface geometry. This extraction of only the necessary portion of the volume data reduces the computational burden while preserving the robustness advantages of volume-based methods.

Inventive Principle:
Principle #2Taking out (Extraction)

2Productivity

If surface-based methods are used for real-time reconstruction, then computational complexity is reduced, but susceptibility to calibration errors and silhouette inaccuracies increases

Engineering Contradiction:
Improvereal-time reconstruction speedVSAvoidsusceptibility to calibration errors
Core Design Contradiction:
ProductivityVSReliability

Solution Approach 1:

The patent segments the reconstruction process into two phases: an initial volume-based phase that processes all voxels to ensure robustness against calibration errors, and a subsequent surface extraction phase that operates on the refined voxel data. This segmentation allows the system to benefit from both volume-based robustness and surface-based efficiency in real-time operation.

Inventive Principle:
Principle #1Segmentation

3Manufacturing precision

If high spatial resolution is achieved in 3D reconstruction, then manufacturing precision is improved, but computational complexity increases cubically

Engineering Contradiction:
Improvespatial resolutionVSAvoidcomputational complexity
Core Design Contradiction:
Manufacturing precisionVSDevice complexity

Solution Approach 1:

The workspace is divided into a 3D grid where only the outer layer of voxels (those potentially containing surface information) are processed in detail, while interior voxels are handled more coarsely or excluded from intensive processing. This segmentation enables high spatial resolution at the surface without the cubic computational cost of processing the entire volume at the same resolution.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent applies different processing qualities to different regions: high-resolution processing is applied only to the near-surface voxels where accurate surface reconstruction is needed, while lower-resolution or simplified processing is applied to interior regions. This local quality differentiation maintains manufacturing precision at the surface while avoiding the cubic complexity increase that would result from uniform high-resolution processing throughout the entire volume.

Inventive Principle:
Principle #3Local quality

Data Source

PatentEP2315181B1Method and system for reconstructing the surface of 3D objects in a multi-camera system
Publication Date: 2015.04.01 DEUTSCHE TELEKOM AG
  • EP2315181B1 patent drawingFigure 1~2
  • EP2315181B1 patent drawingFigure 3~4a
  • EP2315181B1 patent drawingFigure 4b

AI summary

The method involves connecting multiple calibrated and synchronized video cameras (11) via a data network, where a three-dimensional object is located in a three-dimensional working space (12) defined by the cameras. A single-time initial calculation of three-dimensional data in the form of a linearly arranged three-dimensional point stack and imaging data, is performed on basis of an imaging model. A volume-based process for detecting only surface-near volume elements of the object, is used for surface reconstruction. The working space is represented by a grid of volume elements. An independent claim is also included for a system for reconstructing a surface of a three-dimensional object, comprising multiple cameras.