Monocular Volumetric Reconstruction of a Fluid Flow Based on Image Registration and Molecular Tagging

We propose to combine image registration and volumetric reconstruction from a monocular video of a Hele-Shaw (hs) cell being drained. An hs cell is a tank whose thickness is small (e.g. 1 mm) compared to the other dimensions (e.g. 400 × 800 mm2). A pattern is tagged in the liquid at a molecular level by photobleaching [1]. The evolution of the pattern is filmed with a camera whose principal axis is orthogonal to the wall of the cell. Classical methods cannot be used directly to track the pattern because what is observed is the integration of the marked molecules over the entire thickness of the cell and because the velocity of the fluid is not uniform across the thickness of the cell. The proposed approach extends classical direct image registration (dir) [2] by incorporating a volumetric image formation model and simple constraints based on the geometry of the experimental setup. It allows us to accurately measure the motion and the velocity profiles for the entire volume which is something usually hard to achieve. The results we obtain are consistent with the theoretical hydrodynamic behaviour for this flow which is known as the planar Poiseuille flow (ppf). The interest of this work is twofold. In computer vision, it is a new algorithm for monocular volumetric reconstruction. In fluid mechanics, it allows us to experimentally measure the ppf in a narrow confined environment and it is a method based on two new elements: photobleaching and dir. Computer vision algorithms are mainly dedicated to opaque objects. A review of algorithms for transparent objects is given in [3]. A work sharing similarities with our is [4] where the authors proposes a variant of optical flow estimation from images that uses a volumetric model of a liquid in a microchannel. However, contrarily to our approach, [4] uses the ppf as a strong assumption.