Indoor 3D Surface Reconstruction via Building-Object Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing 3D meshing algorithms for building interiors combine objects and structural elements into a single mesh, leading to inaccurate architectural models and storage inefficiencies due to uniform detail representation, which is not faithful to planar elements like floors and walls.
Innovation Solution
The method involves segmenting building and object elements from a building scan using an electronic computing device by generating an octree and 2.5D model, extracting interior and exterior volumes, and meshing them separately to create accurate 3D object and building geometries, allowing for efficient representation of planar elements with fewer triangles and detailed objects with more triangles.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If a single mesh algorithm is used to represent all building elements and objects together, then the representation is simplified and easier to generate, but the accuracy of building structure elements deteriorates due to clutter from objects and uniform detail representation
Solution Approach 1:
The patent segments the 3D space into interior and exterior volumes using an octree data structure, separating building structure elements from objects. This segmentation allows different meshing strategies to be applied to different regions, improving the accuracy of building structure representation while maintaining ease of generation through automated processing.
2Ease of manufacture
If all elements are represented with the same level of detail in a single mesh, then the generation process is simpler, but storage and transmission efficiency deteriorates due to large number of triangles
Solution Approach 1:
The patent applies local quality by representing different elements with appropriate levels of detail. Building structure elements (walls, floors, ceilings) are represented with fewer triangles using planar surface approximation, while objects are represented with higher detail. This reduces the total number of triangles while maintaining visual fidelity where needed.
3Ease of operation
If a single mesh is used to represent both building structures and objects, then the processing is more straightforward, but the fidelity to actual building geometry deteriorates due to clutter from objects
Solution Approach 1:
The patent extracts objects from the building structure by classifying scan points as either building elements or objects based on their spatial relationship to the octree interior/exterior volumes. This extraction allows the building geometry to be reconstructed without object clutter, improving measurement precision while maintaining straightforward processing through automated classification.
Data Source
AI summary
Methods for indoor 3D surface reconstruction and 2D floor plan recovery by segmenting a number of objects and building structure elements from a building scan using an electronic computing device are presented, the methods including: causing the electronic computing device to capture the building scan, where the building scan includes a number of scan points; pre-processing scan data from the building scan; generating an octree and a 2.5D model from the pre-processed scan data; extracting interior and exterior volumes from the octree model and the 2.5D model; and meshing the extracted volumes to generate a 3D object geometry and a 3D building geometry, where the 3D object geometry corresponds with the number of objects and the 3D building geometry corresponds with the indoor 3D surface reconstruction of building structure elements.