Visuals of a europlanet #RASJuno talk about JunoCam image processing, London, 2018-05-10.
The talk essentially investigated the feasibility of 1st order short-term weather forecast derived from a pair of JunoCam images taken within a few minutes.
It starts with a Perijove-12 flyby movie to provide some context. Juno's Perijove-12 flyby has taken place on 2018-04-01.
It then compares cropped pairs of locally contrast-normalized JunoCam images reprojected to the same vantage point, and visualizes according band-pass filtered displacement fields. Morphs extrapolate the motion between the two images of the JunoCam image pair by a factor of 100 into the past, and also into the future, assuming a stable velocity field. The differential equation given by the displacement field was forward and backward integrated using the probably most simple and 1st order numerical method, called Euler method. Numerical integration is the basis of the morphs. Changing velocity fields, like moving storm systems haven't modeled in this feasibility test.
The second image pair is investigated in more detail. After various versions of visualizing the displacement field itself, first and second order derivatives, such as curl, divergence, laplacian, are visualized.
Then statistical errors induced by the specific choice of sets of tiles for stereo correlation are visualized.
This raises the question, whether a larger sample of correlation tiles can smooth the noise. This effect is visualized, too.
Another question is the feasibility of a higher-resolved velocity map by working with smaller tiles, and changing the upper frequency bound of the band-pass filter applied to the raw displacement data. This effect is visualized.
A few morphs are added towards the end of the movie.
The design, implementation, debugging, an test runs of the analysis software, including movie and talk preparation took about two weeks. So, the result might not yet be quite free of remaining glitches.
This investigation was primarily about feasibility. Exploring the limitations of those methods, and applying more sophisticated techniques, as well as the reduction to physical entities like velocity or vorticity is ongoing.
Credit for raw images: NASA / JPL / SwRI / MSSS
Navigation data: NAIF/SPICE
Image processing and data reduction: Gerald Eichstädt
Movie compilation made extensive use of ffmpeg.
MP4 version, about 630 MB: