Simply amazing!
Are you also able to obtain a 3d STL model from those data?

As it is now it is just a heightfield. It is just a quick proof of concept. I am going to use it to build a full shapemodel.

WOW!!!! That's amazing!!! Awesome!!!!
While you build full model can you upload some more of the so far finished stuff....perhaps youtube???

It's possible (and Gerald already did this) but it's complicated by the shape of 67P and it's very time consuming job especially in case of rotation animation (which is possible thanks to the NavCam images).
My guess is at least one hundred control points for every pair of images. For whole animation (one revolution) it's 1300-1500 control points.



It looks very impressive! It reminds me some works from the former(?) UMSF member who did 3D images from HiRISE cameras by shape-from-shading method.

By automated matching of 256x256 = 65536 (overlapping) small tiles, and storing the respective displacements, I've tried to get this somehow managed.
But areas with few features, e.g. large shadows or very smooth patches, are hard to match, and result in very noisy parts, despite noise filtering. Some features aren't visible in both images, so again hard to match.
Here a visualization of such a displacement map:

The displacement map can then be used to generate a radius map (kind of a depth map), provided some telemetry data are known or estimated:

From this, a 3d-mesh could be generated. (But the quality of the above map isn't too convincing.)
I've used the displacement map to generate the interpolated frames.
(Credit for the underlying images, used to generate the above maps: ESA / Rosetta / MPS for OSIRIS Team MPS / UPD / LAM / IAA / SSO / INTA / UPM / DASP / IDA)

down in the -- north up and south down ---
a repost from a different forum
the bottom two images in the bottom right image


been working on mapping the esa jpg images to a updated mesh
-- not looking good YET
a 10 deg grid test map
then warped to the mesh -- this is SCARY
( as in Oct. 31 scary )
then just the updated mesh and esa images from aug. 7 and 15

What I would like to see is an image of the comet surrounded with short arrows pointing in the "down" direction.

with "Down as the direction of the SouthPole
a few images





Navcam from the 14,15,20 and two images of my mesh in blender
Copyright: ESA/Rosetta/NAVCAM
and me - John vanvliet

a DEM from the mesh i am using
the center is 0 long / 0 lat -180 to 180

A similar result would be obtained by color coding a map of the comet to show height in reference to the "geoid”. The colors would quickly show which regions of the comet were uphill or downhill from each other.

Of course, the shape of 67P C-G's geoid will be interesting in itself – and something of a chore to compute.

If you were to take two spherical objects and connect them with a stick (or have them orbit each other) then the low potential point would not be at the midpoint between them, I.e., at the center of mass of the two body system. There would be two low points, at the center of mass of each of the two spheres. The midpoint of the system is a saddle point, with a higher potential than either of the two centers.

Is that your own shapemodel or is it something ESA have made that is in the public domain?

Im working on building my own. but I would like to have another stereo pair from osiris before i continue in ernest. (i want to make an ultra high density model)

I have reverseengineered all the spacecraft positions relative to the comet for all pictures up until now but i have no real coordinate system yet. Is there a consensus northpole and a southpole and a prime meridian that i can use to build that?

Just points on the surface would help


edit:
oh.. you just posted those pictures... where did you get that information?

That's absolutely correct, but materials on the comet surface cannot fall towards either of the two centres. They are constrained to move on or above the surface. That being so, the neck is the lowest potential location they can migrate to.

no it is all mine
i used motion to make a "point cloud" from two videos
then meshed it and just finished a rough edit

i am about to add details from the images using "shape from shade"
an old bit of code from a siggraph paper "mini_sfs"
and a newer phd paper and code
http://code.google.com/p/hyperion-cv/

i have NOT uploaded version TWO yet
version 1 is here
https://googledrive.com/host/0B6ZYAd08tZL-V...21-2014Mesh.zip

--- VERSION 2 is here "67P.09.03.2014.zip"
https://googledrive.com/host/0B6ZYAd08tZL-V....09.03.2014.zip

I am very impressed by this, John! Looking forward to seeing it evolve.

Phil

OK, here's my super simple comet model. Two spheres, radius 2 and radius 1, connected by a cylinder neck of length = 1 and radius = 1/2. Streamlines are down the gradient of the potential. Contours are the potential. YMMV on accuracy. I can improve it and generate the 3D field with some more accurate estimations of the lobes and neck. Fun to fiddle with.