Subgrid scale formulation of optical flow for the study of turbulent flow

Optical flow for incompressible turbulence motion estimation

“…As for the synthetic scalar sequence, the Mixed model along with the novel regularizer, estimates precisely the behavior of subgrid scale motion and its effect to the resolved one (as shown in velocity spectra). The proposed formulation leads to increase the precision by 50% in comparison with Cassisa et al (2011) and outperforms other state-of-the-art optical flow methods in the literature. …”

“…We can clearly observe that the proposed formulation enables to improve the estimation. The vortex structures of different scales, which are comparatively blurred with H&S and are slightly detected by Cassisa et al (2011), are much more accurately estimated with the subgrid modeling, such that the vortices have a similar behavior to that of DNS .…”

“…However, when it comes to turbulent flow, the interaction between subpixel scales (unresolved scales) and resolved ones plays an important role in energy transport, which should be taken into account. Recently, Cassisa et al (2011) and Zille et al (2014) introduced the concept of Large Eddy Simulation (LES), by considering the motions under pixel as subgrid turbulent dissipation. However, the coefficient of turbulent dissipation is computed a priori from velocity spectrum, which is not practically accessible since the velocity field is unknown neither in physical space nor in spectral one.…”

“…Results of the proposed formulation (solid line) are compared to H&S (dotted line) and Cassisa et al (2011) (dashed line). It is visible that the introduction of sub-pixel motion modeling (Cassisa et al 2011) has already improved the precision of estimation by 40% for AAE and 34% for RMSE. However, even compared to this high-performance method, our new formulation leads to a further improvement by 50% approximatively for both criteria.…”

Comparison between optical flow and cross-correlation methods for extraction of velocity fields from particle images

“…Although the first-order Tikhonov constraint functional is physically plausible, it is not derived from the first principles. A recent effort has been made by incorporating certain physical mechanisms in turbulent flows into the constraints in optical flow computations (Cassisa et al 2011;Zille et al 2014;Chen et al 2015).…”