Tracer-based 3D surface reconstruction

Abstract

The continuous advancements of particle imaging techniques for flow field measurements have led to imaging systems and processing approaches matching the demands for 3D velocimetry at large scale (Schanz et al., 2016; Discetti and Coletti, 2018). Often, the flow past an object immersed in the fluid is of key interest, and in some cases the experimentalist exploits the velocimetry data for analysis of the near-surface flow properties such as pressure. It follows that knowledge of the object shape and position is essential. For 2D studies, the issue of identifying the fluid-solid interface often reduces to detection of the intensity gradient resulting from the light sheet striking the object. The latter task is well explored, with a variety of methods providing the object interface in the measurement plane (Canny, 1986; Malik et al., 2001; Gui et al., 2003, among others). These approaches, however, are not applicable in volumetric studies where the illumination is diffuse. A frequently applied alternative in fluid-structure-interaction studies is a dual-measurement approach, where a second measurement system tracks the object shape (e.g., Acher et al., 2019; Zhang et al., 2019) The complexity of operating two measurement systems may not be affordable however, motivating the development of 3D interface detection methods that rely solely on the flow imaging system. Particle imaging based interface detection approaches in 3D have been addressed from various perspectives, containing the detection of fluid-fluid interfaces. Examples utilizing tomographic PIV measurements include the studies of Adhikari and Longmire (2012), Im et al. (2014), and Ebi and Clemens (2016). The two latter examples identify the fluid-solid interface by discriminating a seeded phase (the fluid) from a void phase (the solid). The present work is inspired by this principle, but it assumes a discrete 3D particle distribution as obtained from a generic particle tracking algorithm such as IPR by Wieneke (2012), or “Shake-The-Box” by Schanz et al. (2016), as a foundation. Summarizing, this work aims to detect the surface of a solid object immersed in a seeded flow, solely based on the spatial distribution of flow tracers as recorded by a generic 3D PTV measurement.