A comparison of state-of-the-art image evaluation techniques for analysis of opaque flows

Abstract

In this paper we compare state-of-the-art fluid motion estimation algorithms with respect to the application to opaque flow images, such as Particle Image Velocimetry (PIV), Optical Flow and specialized variants of these. In order to compare the outcome of the algorithms we perform turbulent jet experiments with different Reynolds numbers to generate opaque images with a known average flow field as a reference to estimate the errors. Thereafter the algorithms are used to estimate the velocity of a submerged oil/gas plume that was recorded by an observation camera. We will show that the estimates strongly depend on the choice of algorithm, especially for field test images where cloud-like image structures are the only information to retrieve the motion.