Identification of liquid–vapor phase transition using the co-occurrence matrix and Haralick features
Résumé
Identification of a phase transition — specifically, the precise timing of the critical temperature crossing — is challenging due to the thermal inertia of the enclosure holding the fluid, the fluid’s finite thermal conductivity, convective flows, and sedimentation. The human eye is particularly well adapted to detecting such phase transitions in images produced by transmitted light through critical fluids. In this study, we used Haralick image features that most closely correspond to human visual perception to identify gas–liquid phase transitions. The images were recorded during the phase transition of sulfur hexafluoride (SF ) in microgravity—a particularly challenging scenario. The first goal of this study was to demonstrate that Haralick features can identify a phase transition from recorded images of supercritical fluids in microgravity. While highly sensitive to minute changes in image detail, Haralick features are not inherently normalized; thus, conclusions based on their values for a given image bit depth cannot be reliably compared to those derived from images captured with different cameras or quantization schemes. The second goal of this study was to identify empirical scaling laws for Haralick features that enable the prediction of their behavior as image bit depth varies. Such flexibility would allow direct comparison of results obtained from phase transition datasets recorded using different quantization schemes and imaging systems.