Motion tracking-enhanced MART for tomographic PIV

Abstract

A novel technique to increase the accuracy of multiplicative algebraic reconstruction technique (MART) reconstruction from tomographic particle image velocimetry (PIV) recordings at higher seeding density than currently possible is presented. The motion tracking enhancement (MTE) method is based on the combined utilization of images from two or more exposures to enhance the reconstruction of individual intensity fields. The working principle is first introduced qualitatively, and the mathematical background is given that explains how the MART reconstruction can be improved on the basis of an improved first guess object obtained from the combination of non-simultaneous views reduced to the same time instant deforming the 3D objects by an estimate of the particle motion field. The performances of MTE are quantitatively evaluated by numerical simulation of the imaging, reconstruction and image correlation processes. The cases of two or more exposures obtained from time-resolved experiments are considered. The iterative application of MTE appears to significantly improve the reconstruction quality, first by decreasing the intensity of the ghost images and second, by increasing the intensity and the reconstruction precision for the actual particles. Based on computer simulations, the maximum imaged seeding density that can be dealt with is tripled with respect to the MART analysis applied to a single exposure. The analysis also illustrates that the maximum effect of the MTE method is comparable to that of doubling the number of cameras in the tomographic system. Experiments performed on a transitional jet at Re = 5000 apply the MTE method to double-frame recordings. The velocity measurement precision is increased for a system with fewer views (two or three cameras compared with four cameras). The ghost particles' intensity is also visibly reduced although to a lesser extent with respect to the computer simulations. The velocity and vorticity field obtained from a three-camera reconstruction with MTE are equivalent to that from a four-camera analysis. Possible variants of the MTE algorithm are investigated based on a first guess obtained by average or by product of pseudo-simultaneous objects (PSO), which potentially offer a higher convergence rate.