Panoramic Indoor Spatial Modeling with Boundary-Based Point Cloud Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for creating three-dimensional spatial models from panoramic images are time-consuming, costly, and result in resolution loss or vertex density issues, particularly in complex or elongated spaces, leading to holes or increased model size.
Innovation Solution
A server-based method that derives boundary point information from a panoramic image, creates a point cloud, divides it based on these points, projects the cloud onto planes to form a texture image, extracts geometric information to build a mesh model, and constructs a spatial model using the texture and mesh.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If all images are normalized to increase processing speed, then spatial analysis speed is improved, but image resolution is lost and vertex density decreases in complex or elongated spaces
Solution Approach 1:
The patent divides the panoramic image into multiple plane images (ceiling, floor, walls) based on boundary point information. Each plane image is then processed separately to create corresponding plane point clouds and texture images. This segmentation allows the system to maintain high resolution in each plane while managing processing complexity, resolving the contradiction between speed and resolution by processing smaller, specialized images rather than normalizing the entire large image.
Solution Approach 2:
The patent applies different processing strategies to different regions (planes) of the panoramic image. Each plane receives customized processing based on its specific characteristics and boundary point information, allowing high-resolution texture extraction where needed while optimizing processing speed. This local quality approach ensures that resolution is maintained in critical areas without uniformly processing the entire image at high cost.
2Productivity
If normalized images are used to create point cloud model, then processing speed is increased, but vertex density decreases causing holes or increased model size in complex spaces
Solution Approach 1:
The patent segments the point cloud creation process by plane, creating separate plane point clouds from plane-specific texture images and boundary point information. This segmentation enables the system to generate dense vertex distributions in each plane without needing to process the entire normalized image, thereby maintaining manufacturing precision while improving overall productivity through parallel processing of multiple planes.
Solution Approach 2:
The patent introduces boundary point information as an intermediary element that connects the panoramic image to the plane point clouds. This intermediary allows the system to accurately transfer geometric and textural information from the original high-resolution panoramic image to the plane-specific representations, ensuring high vertex density and preventing holes in the final model without requiring uniform normalization of all images.
3Manufacturing precision
If expert manually creates three-dimensional spatial model, then model accuracy is improved, but time and cost increase significantly
Solution Approach 1:
The patent implements an automated system that processes panoramic images to generate three-dimensional spatial models without requiring manual intervention from experts. The system automatically extracts boundary point information, creates plane point clouds, generates texture images, and constructs the final spatial model. This self-service approach maintains high model accuracy through automated algorithms while dramatically reducing the time and cost associated with manual expert creation.
Solution Approach 2:
The patent replaces the mechanical manual process of expert modeling with an automated computational system. Instead of manual creation, the system uses image processing algorithms, point cloud generation, and mesh construction techniques to automatically create accurate three-dimensional spatial models. This substitution of mechanical manual work with automated digital processing achieves both high accuracy and reduced time consumption.
Data Source
AI summary
A server that creates a spatial model includes a derivation unit configured to derive boundary point information from a panoramic image of an indoor space; a point cloud creation unit configured to create a point cloud for the panoramic image; a division unit configured to divide the point cloud based on the boundary point information; a texture image creation unit configured to create a texture image by projecting the divided point cloud onto a plane corresponding to the divided point cloud; a mesh model creation unit configured to extract geometric information from the point cloud based on the boundary point information, and create a mesh model based on the extracted geometric information; and a spatial model creation unit configured to create the spatial model for the indoor space based on the texture image and the mesh model.