Architectural Layout Determination via MRF Energy Minimization

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current methods for determining architectural layouts in 3D reconstructed indoor scenes are limited by noise in data, reliance on specific assumptions about layout shapes, occlusions, and sensor positions, and require color data or machine learning, making them prone to errors and incomplete reconstructions.

Innovation Solution

A computer-implemented method using a Markov Random Field (MRF) energy minimization to determine architectural layouts by assigning direction normals to points representing walls, which reduces angular distance penalties and identifies maximal sets of consecutive points to define wall segments, allowing for robust and precise layout extraction without assumptions on shape or sensor position, and without requiring color data or machine learning.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If machine learning methods are used for scene understanding, then classification and segmentation accuracy may improve, but computational complexity and data requirements increase

Engineering Contradiction:
Improveclassification accuracyVSAvoidcomputational complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent replaces machine learning approaches with a geometric constraint-based system. Instead of using neural networks or probabilistic models that require extensive training data and computation, the invention uses deterministic geometric relationships (parallelism, perpendicularity, coplanarity) to classify points and reconstruct layouts, significantly reducing computational complexity while maintaining accuracy

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The system uses the geometric structure inherent in the point cloud data itself to solve the classification problem. By exploiting the natural geometric constraints of architectural layouts (walls being parallel/perpendicular, floors and ceilings being coplanar), the method makes the data self-descriptive without requiring external training datasets or complex models

Inventive Principle:
Principle #25Self-service

2Measurement precision

If color data is used for scene understanding, then object recognition accuracy improves, but data processing time and storage requirements increase

Engineering Contradiction:
Improveobject recognition accuracyVSAvoiddata processing time
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The patent extracts and discards the color component from RGB-D data, utilizing only the geometric depth information. By removing the color channel, the system reduces data processing time and storage requirements while maintaining sufficient accuracy for architectural layout reconstruction through geometric constraint satisfaction

Inventive Principle:
Principle #2Taking out (Extraction)

Solution Approach 2:

The invention changes the relevant parameters from including color information (RGB-D) to excluding color information (D-only). This parameter change simplifies the data representation and processing pipeline, focusing computational resources on geometric relationships that are sufficient for layout reconstruction

Inventive Principle:
Principle #35Parameter changes

3Productivity

If assumptions about layout shapes are made, then reconstruction speed increases, but accuracy and adaptability decrease

Engineering Contradiction:
Improvereconstruction speedVSAvoidlayout accuracy
Core Design Contradiction:
ProductivityVSMeasurement precision

Solution Approach 1:

The patent implements a dynamic constraint satisfaction approach where geometric constraints are applied adaptively during the reconstruction process. Rather than assuming a fixed layout shape beforehand, the system dynamically identifies and applies relevant geometric constraints (parallelism, perpendicularity, coplanarity) based on the actual point cloud data, allowing both speed and accuracy

Inventive Principle:
Principle #15Dynamics

4Loss of information

If complete scene understanding is performed, then semantic information increases, but computational resources and processing time increase

Engineering Contradiction:
Improvesemantic informationVSAvoidcomputational resources
Core Design Contradiction:
Loss of informationVSDevice complexity

Solution Approach 1:

The patent segments the scene understanding task into distinct geometric classification categories (floor, ceiling, wall, other vertical) based on normal vector analysis. This segmentation allows the system to extract sufficient semantic information for architectural layout reconstruction without performing complete general scene understanding, reducing computational resource requirements while maintaining necessary semantic content

Inventive Principle:
Principle #1Segmentation

Data Source

PatentEP3401815B1Determining an architectural layout
Publication Date: 2022.12.21 DASSAULT SYSTEMES SA
  • EP3401815B1 patent drawingFigure 1~2
  • EP3401815B1 patent drawingFigure 3
  • EP3401815B1 patent drawingFigure 4~6

AI summary

The invention notably relates to a computer-implemented method for determining an architectural layout. The method comprises providing a cycle of points that represents a planar cross section of a cycle of walls, and, assigned to each respective point, a respective first datum that represents a direction normal to the cycle of points at the respective point. The method also comprises minimizing a Markov Random Field energy thereby assigning, to each respective point, a respective one of the set of second data. The method also comprises identifying maximal sets of consecutive points to which a same second datum is assigned, and a cycle of vertices bounding a cycle of segments which represents the architectural layout. Such a method constitutes an improved solution for determining an architectural layout.