Treffer: 基于自适应八叉树的深度学习地质建模 并行显式化采样方法.

In the explicit modeling phase of deep learning-based geological modeling techniques traditional regular grid sampling methods often encounter significant computational and storage demands. They frequently produce aliasing effects at model boundariesmaking it difficult to balance efficiency with visualization accuracy. To overcome these limitations, this paper proposes a parallel explicit sampling method for deep learning-driven geological modeling based on an adaptive octree structure. The proposed approach involves three key steps: training a deep learning classification network on borehole data to capture the spatial distribution patterns of subsurface lithology; employing an adaptive octree data structure to automatically refine nodes in regions identified as lithological interfaces enabling high-resolution sampling in critical zones while maintaining coarse resolution in homogeneous areas to minimize redundant triangular facets; and implementing a multi-process parallel computing framework to enhance processing efficiency for large-scale data. Experimental results indicate that, compared to conventional regular grid-based explicit methods, the proposed approach reduces the number of model facets by over 60% without compromising modeling accuracy, while markedly improving boundary smoothness. This method effectively alleviates rendering performance bottlenecks and enhances the precision of model visualizationproviding a robust solution for efficient and accurate geological modeling. [ABSTRACT FROM AUTHOR]

在深度学习地质建模技术的显式化过程中, 传统规则栅格采样方式面临巨大的计算与存储压力, 且常出现模型边界的 锯齿效应, 难以兼顾效率与可视化精度。为此, 该文提出一种基于自适应八叉树的深度学习地质建模并行显式化采样方法。 首先, 基于钻孔数据训练深度学习网络, 利用训练好的深度学习网络学习地下岩性的空间分布规律;其次, 基于自适应八叉树 的数据结构, 在识别到岩性界面区域时自动对相应爷点进行细化分裂, 从而在关键部位采用高分辨率采样、在均质区域采用 粗分辨率采样以减少不必要的三角面数;最后, 通过多进程并行计算框架提高该方法对大规模数据的处理效率。实验结果表 明, 相比基于规则栅格的常规显式化方法, 该方法在保证建模精度的前提下, 模型面数减少60 %以上, 且模型边界光顺性明显 提升, 在有效緩解渲染性能压力的同时, 能提高模型可视化的精细度。. [ABSTRACT FROM AUTHOR]

Copyright of Geography & Geographic Information Science is the property of Geography & Geo-Information Science Editorial Office and its content may not be copied or emailed to multiple sites without the copyright holder's express written permission. Additionally, content may not be used with any artificial intelligence tools or machine learning technologies. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)