Analysis of Slush-State Crystal Structures through Shape-From-Silhouettes 3D Reconstruction

This work proposes a novel approach for reconstructing the three-dimensional (3D) shapes of fluid crystal formations in their slush state, characterised by the coexistence of liquid and solid phases at the triple point temperature. Slush mixtures present higher density and lower specific enthalpy compared to their liquid counterpart, thus representing attractive candidates for different applications, such as next-generation propellants in space and ground transportation. However, the presence of solid particles in the liquid phase poses notable obstacles in accurately capturing their complex morphologies, and innovative strategies are required to efficiently characterise and employ slush propellants. The present study leverages the Visual Hull technique, which is a well-known method in computer vision and graphics, to address this challenge. The objectives of this research are threefold: to evaluate the efficacy of advanced computer vision methodologies in biphasic fluid characterisation; to provide a better understanding of crystal structure formation within the slush state; and lastly, to produce an extensive database, encompassing different types of fluid, which will serve as foundation for validating predictive models. Each of these activities constitutes a benchmark in developing groundbreaking storage methodologies for new propellants. The first part of the work consists of a comprehensive study of several fluids to identify samples that exhibit similar features to real next-generation propellants (in particular hydrogen) and define a variety of test cases. The Visual Hull is then presented and explained, along with the related algorithm for developing the 3D model. Subsequently, a feasibility study is conducted to determine the minimum number of cameras that could be used while maintaining an acceptable error in the reconstruction. The following part describes the experiment implementation. A dedicated facility was designed and built, consisting of 8 paired cameras positioned on a circle around a transparent tank, filled with the chosen test fluid in its liquid phase. From above, a frozen body is suspended in the tank. The experimental section showcases the application of the Visual Hull technique to multiple scenarios involving different samples under controlled conditions. The results obtained from the 3D reconstructions are presented and the performance of the algorithm is evaluated. In the final section, the broader implications of this research are discussed and future research directions are delineated. Overall, this thesis provides a solid contribution to the advancement of computational imaging of slush-state fluid crystals, offering a holistic framework for precise 3D shape reconstruction. Moreover, the integration of Visual Hull-based methodologies in biphasic fluids observations represents a noteworthy leap for gaining insights into the qualities of this fascinating state.