Synthetic particle pack generation for augmentation and testing in geological tomographic segmentation

Abstract

3D imaging of granular packings and geological particle samples by computed tomography offers the means for non-destructive analysis. However, obtaining such tomograms with the corresponding segmentation labels, i.e. a unique label per particle, remains a significant challenge. This study introduces a novel physics-based simulation workflow that generates synthetic tomograms with corresponding ground truth segmentations. The synthetic dataset generation tool produces realistic particle pack tomograms in large quantities, supporting data augmentation and serving as a benchmark for geological tomographic segmentation testing. The code in this study is publicly available at: github.com/bogongwang/particle-pack-generation.