3D Voxel Plane Inference for Watertight Virtual Manifold Generation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for deriving information from images and depth information to recreate structures are incomplete, leading to missing or occluded areas in virtual representations, which affects the accuracy of floor plans and physical simulations.
Innovation Solution
The method involves using camera images with tracked location and orientation to generate depth images, inferring missing information through semantic labeling and plane intersection analysis, and creating a complete and watertight virtual wrapper or manifold, which includes structural and non-structural elements.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If traditional image capture methods are used to recreate structures, then the process is simple, but the completeness and accuracy of the virtual representation deteriorates due to missing and occluded areas
Solution Approach 1:
The system performs preliminary actions by capturing images from multiple viewpoints before reconstruction, and by pre-processing images to identify and mask occluding objects. This preliminary capture and analysis enables complete structure reconstruction without requiring physical removal of furniture during scanning.
Solution Approach 2:
The system transitions from 2D images to 3D virtual representations by capturing images from multiple viewpoints and using depth information. This dimensional transformation allows reconstruction of complete structures including areas occluded in any single 2D view, resolving the completeness issue.
2Manufacturing precision
If multiple images are captured to improve completeness, then the accuracy of virtual representation improves, but the time and complexity of the scanning process increases
Solution Approach 1:
The system captures more images than strictly necessary (excessive action) to ensure all structural elements are visible from at least one viewpoint. This redundancy guarantees completeness while the automated processing efficiently handles the increased data volume, maintaining practical scanning times.
Solution Approach 2:
The system creates multiple copies of the scene from different viewpoints through image capture, then processes these copies to reconstruct the complete 3D structure. This copying approach enables accurate floor plan generation without requiring physically exhaustive scanning procedures.
3Manufacturing precision
If occluding objects are removed to improve image quality, then the completeness of structural information improves, but the ease of operation deteriorates
Solution Approach 1:
The system extracts and masks occluding objects from images through automated object recognition and segmentation. By digitally removing these obstructions during processing rather than physically removing furniture during scanning, the system maintains completeness of structural information while preserving ease of operation.
Solution Approach 2:
The system introduces an intermediary processing stage that identifies occluding objects and masks them from the reconstruction process. This intermediary layer allows the scanner to operate easily without moving furniture, while still achieving complete structural information by excluding occluding objects from the 3D reconstruction calculations.
Data Source
AI summary
The described implementations relate to images and depth information and generating useful information from the images and depth information. One example can identify planes in a semantically-labeled 3D voxel representation of a scene. The example can infer missing information by extending planes associated with structural elements of the scene. The example can also generate a watertight manifold representation of the scene at least in part from the inferred missing information.