An Investigation of 3D Sand Particle Fragment Reassembly

Potential fracture surface propagation is an essential characteristic in determining the mechanical properties of natural sands. This paper presents a novel method of obtaining the realistic location and scale of fracture surfaces from 3D particle fragment reassembly by using a self-designed mini-loading apparatus combined with X-ray micro-computed tomography to collect data. The 2D images were processed by de-noising and watershed segmentation algorithms to reconstruct 3D broken particles and obtain information on the separated fragments. Based on the degree of mean curvature at every point in the curve, the boundary of the fragment was extracted and connected by a minimum spanning tree (MST) algorithm. Because of the existing short branches, bottom-up graph pruning was adopted to prune the MST. To improve work efficiency, simple chordless cycles were introduced to separate the boundary of the fragment into several parts. We used a modified 4-points congruent set algorithm to acquire the key points and regarded the descriptor vector as the feature representation. The Möller-Trumbore algorithm was used to detect whether the matched points lay within the triangulated volume to avoid incorrect matching between two fragments. The matching results showed that the proposed approach is capable of reassembling fractured particles and is conducive to better prediction of the area of potential breakage.