Adaptive 3D Mesh Reconstruction for Noisy Sparse Scan Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing mesh reconstruction methods suffer from low accuracy due to data noise, missing data, and equipment precision limitations, leading to large noise in sparse areas and inability to adaptively reconstruct meshes of different smoothness, resulting in inefficient and memory-intensive models.
Innovation Solution
A mesh reconstruction method that presets multiple mesh resolutions, constructs an adaptive octree based on scanning data, calculates node information, determines resolution levels, and adjusts scalar field calculations to extract multi-resolution meshes, reducing noise and improving accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional mesh reconstruction methods are used, then the reconstruction process is simple, but the accuracy of the reconstructed mesh model is low due to noise and data missing
Solution Approach 1:
The patent divides the reconstruction space into a hierarchical octree structure with multiple levels, where each node represents a spatial region. This segmentation allows different resolution levels to be applied to different areas, improving accuracy in critical regions while maintaining simplicity in less important areas, thereby resolving the contradiction between accuracy and complexity.
Solution Approach 2:
The patent implements adaptive mesh refinement by determining local quality metrics for different regions of the scanned object. High-quality reconstruction is applied to areas with important features or high curvature, while lower quality is acceptable in flat or less important regions. This local quality approach improves overall accuracy without uniformly increasing complexity across the entire model.
2Measurement precision
If uniform high-resolution mesh is used throughout the model, then the accuracy is improved, but the model scale and memory consumption increase
Solution Approach 1:
The octree structure segments the mesh into hierarchical levels, allowing the system to store and process only the necessary level of detail for each region. Coarse regions use fewer polygons while fine regions use more, optimizing the quantity of mesh data while maintaining accuracy where needed.
Solution Approach 2:
The patent implements dynamic resolution adjustment where the mesh resolution is not fixed but adapts based on local geometric features, curvature, and importance. This dynamic approach allows the model to use high resolution only where necessary, reducing overall model scale and memory consumption while maintaining accuracy in critical areas.
3Measurement precision
If noise filtering is applied to reduce data noise, then the accuracy is improved, but the processing time and computational resources increase
Solution Approach 1:
The patent performs preliminary noise filtering and data processing during the octree construction phase, organizing cleaned data into the hierarchical structure before actual mesh generation. This preliminary action reduces the computational burden during the reconstruction phase, improving accuracy while managing processing time efficiently.
Solution Approach 2:
By segmenting the processing into hierarchical octree levels, the patent applies noise filtering and refinement operations only at necessary levels and in necessary regions. This segmented approach reduces overall processing time compared to applying uniform high-level filtering across the entire dataset, while still improving accuracy where needed.
Data Source
AI summary
The present disclosure relates to a mesh reconstruction method, a device and a storage medium. The mesh reconstruction method includes constructing a data structure according to three-dimensional scanning data and a preset mesh side length. Node information of a target tree node among a plurality of tree nodes included in the data structure is determined. A level of a mesh to be extracted of the target tree node is determined according to the target data. Once a scalar field of the target tree node is calculated according to the node information of a neighboring tree node corresponding to the target tree node and the node information of the target tree node, a reconstructed mesh model is obtained according to the scalar field.