An Experimental 2D–3D Dynamic Image Analysis Framework for Particle Shape Characterization and Morphological Analysis of Lunar Regolith Simulants in Multi-Dimensional Morphospaces

As human lunar exploration advances through NASA’s Artemis mission and ambitions for a permanent lunar presence grow, understanding lunar regolith is increasingly important. Particle shape plays a pivotal role in governing the behaviour of granular materials, affecting regolith strength, angle of repose, packing density, and interactions with landing spacecraft. Quantitative characterization of lunar regolith particles is therefore essential for mission planning and for the development and validation of adequate simulants used in engineering studies and equipment testing.             

Previous studies have therefore investigated various shape properties of lunar regolith samples and their corresponding simulants using both 2D and 3D techniques. While 2D approaches such as  dynamic (DIA) and static image analysis (SIA) are simple and effective, they do not capture the full 3D geometry of particles and are sensitive to viewing orientation. In contrast, 3D approaches such as laser scanning or X-ray microcomputed tomography (µCT) provide high geometric accuracy but are time-intensive, laborious, and computationally demanding, resulting in a limited number of studies performing 3D shape characterization of lunar regolith simulants. More recently, 3D dynamic image analysis (3D-DIA) has emerged as an intermediate approach, approximating 3D particle geometry from multiple projections. However, only a few setups currently exist, and most rely on proprietary software, limiting transparency, reproducibility, and accessibility.             
Furthermore, extracted shape properties are often analysed individually, overlooking the inherently multi-dimensional nature of particle morphology. Emerging quantitative frameworks, such as morphospaces, are therefore needed to comprehensively capture particle shape and enable systematic, holistic comparison across simulants.

To address the challenge of transparent and reproducible 3D shape characterization of granular particles, we present a novel, low-cost 3D-DIA setup paired with an open-source processing pipeline, which incorporates deep learning–based particle detection and a custom tracking algorithm. The accuracy of derived 3D particle shape descriptors is evaluated against high-resolution µCT scans. Building on the recent introduction of bivariate morphospaces for comprehensive particle shape characterization, we extend this framework by including intermediate-scale particle roundness, thereby establishing a trivariate morphospace that captures all shape properties of powder materials obtainable from imaging data. Distributional patterns within these morphospaces are captured using multi-dimensional Gaussian kernel density estimation (KDE), facilitating quantitative comparison between particle populations via density difference mapping. To further support quantitative assessment across simulants, we introduce the morphological richness (MRic) metric, which condenses the overall morphological diversity of a given simulant into a single scalar value.

To evaluate the proposed framework, 3D particle shape descriptors derived from the 3D-DIA setup were compared with reference µCT measurements. The results show strong agreement and substantial improvement over approximations obtained from single-projection approaches using 2D-DIA and 2D-SIA. Multi-dimensional KDE-based morphospace analysis of EAC-1A, JSC-2A, and NUW-LHT-5M reveals distinct differences in particle shape distributions, further quantified by the MRic metric. These findings demonstrate that the proposed approach provides a robust, reproducible, and scalable method for comprehensive characterization of lunar regolith simulant morphology, supporting the design of more representative simulants and enabling improved understanding of material behaviour in future lunar missions and surface operations.