Hierarchical Mesh Dataset Conversion via Voxel Density Maps
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for creating hierarchical-level-of-detail (HLOD) structures for massive mesh datasets are inefficient due to the large amount of data that needs to be loaded and processed, often requiring multiple streaming passes and being computationally expensive, especially when generating sub-hierarchies and efficient bounds, which can necessitate streaming the entire source dataset multiple times despite most data being useless to small areas.
Innovation Solution
The method involves generating a voxel density map, subdividing it into a hierarchy of bounding boxes, determining leaf tiles, and generating ancestor tiles, which allows for adaptive data processing without relying on spatial locality, improving data coherence and enabling parallel processing.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If current algorithms (vertex clustering) are used to generate HLOD structures, then constant memory usage is achieved, but many streaming passes over the entire dataset are required
Solution Approach 1:
The patent divides the massive mesh dataset into multiple partitions or chunks that can be processed independently. Instead of streaming the entire dataset multiple times, the system segments the data spatially and processes only relevant partitions for each tile generation, reducing the number of streaming passes while maintaining constant memory usage.
Solution Approach 2:
The patent performs preliminary spatial indexing and data organization before the actual HLOD generation. By pre-processing the dataset to create spatial structures (such as bounding volume hierarchies or grid-based indexes), the system enables faster access to relevant data during tile generation, reducing the need for repeated full dataset streams.
2Manufacturing precision
If adaptive algorithms (strict k-d tree) are used to generate efficient bounds, then accurate tile boundaries are achieved, but computational expense increases significantly
Solution Approach 1:
The patent uses simplified, computationally inexpensive bounding volume representations (such as axis-aligned bounding boxes or coarse spherical bounds) instead of strict k-d trees. These approximate bounds are sufficient for view culling purposes and can be generated and discarded efficiently without the high computational cost of adaptive spatial partitioning structures.
Solution Approach 2:
The patent changes the parameters of the bounding volume representation from strict adaptive structures to fixed-grid or hierarchy-based structures. By using predetermined grid sizes or hierarchical levels that match the HLOD structure, the system avoids the computational expense of data-adaptive algorithms while maintaining sufficient accuracy for rendering decisions.
3Reliability
If the entire source dataset is streamed multiple times to generate sub-hierarchies, then complete coverage is achieved, but data redundancy increases
Solution Approach 1:
The patent applies local quality by processing only the relevant portions of the dataset needed for each specific tile or region. Instead of streaming the entire dataset uniformly, the system identifies and processes only the local data required for each tile's level of detail generation, reducing redundant data transmission while ensuring complete coverage of necessary areas.
Data Source
AI summary
Systems, methods, devices, and non-transitory media of the various embodiments enable converting massive mesh datasets that may carry a single material to a hierarchical format. Various embodiments may provide processing efficiency and scalability in creating hierarchical format representations of massive mesh datasets and/or in rendering massive mesh datasets.