 |
Center for Nonlinear Studies |
A large number of applications, from satellite imaging and remote sensing to machine learning to video compression, require the ability to extract dynamics from image sequences. With collaborators at the University of California, Berkeley, I have developed a software tool, Advection Corrected Correlation Image Velocimetry (ACCIV), that can extract the motion of tracers in an image sequence to produce a smooth-fit velocity field. ACCIV draws from Correlation Image Velocimetry (CIV), a common laboratory technique for producing velocities from passive tracers through linearly transformed correlations. We have added to ACCIV the ability to track feature motion that is highly non-linear. Thus, we are able to make use of image sequences separated by relatively long periods of time, decreasing the uncertainty in our velocity fields. ACCIV has been developed to construct velocity maps of Jupiter’s and Saturn’s upper atmospheres. We have used these velocity fields constrain dynamical models of vortices and zonal jets that have deepened our understanding of climate cycles on these planets. Recently, I have begun to apply the software to satellite images of glacier dynamics. In this talk I discuss both the ACCIV algorithm and its various applications.
Host: Peter Loxley, loxley@lanl.gov