Towards 3D Shape Estimation from 2D Particle Images: A State-of-the-Art Review and Demonstration

Abstract

Particle shape plays a critical role in governing the properties and behavior of granular materials. Despite advances in capturing and analyzing 3D particle shapes, these remain more demanding than 2D shape analysis due to the high computational costs and time-consuming nature of 3D imaging processes. Consequently, there is a growing interest in exploring potential correlations between 3D and 2D shapes, as this approach could potentially enable a reasonable estimation of a 3D shape from a 2D particle image, or at most, a couple of images. In response to this research interest, this study provides a thorough review of previous studies that have attempted to establish a correlation between 3D and 2D shape measures. A key finding from the extensive review is the high correlation between 2D perimeter circularity (c_p) and Wadell’s true sphericity (S) defined in 3D, suggesting that a 3D shape can be estimated from the c_p value in terms of S. To further substantiate the correlation between c_p and S, this study analyzes approximately 400 mineral particle geometries available from an open-access data repository in both 3D and 2D. The analysis reveals a strong linear relationship between S and c_p compared with other 2D shape descriptors broadly used in the research community. Furthermore, the limited variance in c_p values indicates that c_p is insensitive to changes in viewpoint, which indicates that fewer 2D images are needed. This finding offers a promising avenue for cost-effective and reliable 3D shape estimation using 2D particle images.