Evaluation of a wavelet-based optical flow algorithm through the use of large eddy simulations

Kyle Rocha-Brownell (1) , Pierre Dérian (2) , David H Richter (3) , Peter P Sullivan (4) , Shane D Mayor (1)

1 Atmospheric Lidar Research Group
2 FLUMINANCE - Fluid Flow Analysis, Description and Control from Image Sequences
3 UND - University of Notre Dame [Indiana]
4 NCAR - National Center for Atmospheric Research [Boulder]

Recent progress in observations of wind fields by scanning aerosol lidar and motion estimation algorithms elevates interest in accuracy, precision , and resolution. Toward this, we are using large eddy simulations to advect a passive tracer in the model domain to mimic atmospheric aerosol. A wavelet-based optical flow algorithm is applied to the model output in order to derive dense motion vector fields. Each motion vector is compared with the known wind velocity at the corresponding point.

Mots clés: optical flow; motion estimation; turbulence; wavelet transforms; large eddy simulation; LES; lidar; aerosol; remote sensing

Domaines: Vision par ordinateur et reconnaissance de formes [cs.CV]; Océan, Atmosphère; Mécanique des fluides [physics.class-ph]; Modélisation et simulation

Kyle Rocha-Brownell, Pierre Dérian, David H Richter, Peter P Sullivan, Shane D Mayor. Evaluation of a wavelet-based optical flow algorithm through the use of large eddy simulations. 28th International Laser Radar Conference, Jun 2017, Bucarest, Romania. ⟨hal-01573724⟩