Two-Pulse 3D particle tracking with Shake-The-Box

Abstract

The problem of reconstructing the three-dimensional position of particle flow tracers from their projection on multiple cameras lies at the heart of several 3D particle-image-based velocimetry and Lagrangian particle tracking (LPT) measurement techniques. While cross-correlation-based techniques such as tomographic-PIV (Tomo-PIV, Elsinga et al. 2006) make use of algebraic methods (e.g. MART, Herman and Lent 1976) to reconstruct particles as intensity peaks in a discretized voxel space, triangulation-based methods (3D-PTV, Nishino et al. 1989, Maas et al. 1993 and Iterative Particle Reconstruction, IPR, Wieneke 2013, Jahn et al. 2021) leverage epipolar geometry to reconstruct individual particles as positions and peak-intensity values in the 3D domain. Despite the inherent differences concerning accuracy, robustness and computational cost between the two approaches, in both cases the 3D reconstruction represents a bottleneck when the spatial resolution (i.e. particle image density, indicated in particles per pixel, ppp) of the measurement is considered. In fact, as the number of particles to be reconstructed increases (assuming constant properties of the imaging system), the reconstruction process become increasingly difficult due to the underdetermined nature of the problem. This typically results in a lower number and positional accuracy of the reconstructed particles, as well as an increasing number of spurious particles (ghost particles, Elsinga et al. 2011) which affect the accuracy of the measurement. As a consequence, during the last decade, several methods have been developed to increase the performances of the reconstruction technique; an overview of these developments can be found in Scarano 2012 and Jahn et al. 2021.