From drbitboy@gmail.com Wed Mar 6 17:59:07 2013 Date: Wed, 6 Mar 2013 17:59:06 -0500 From: Brian Carcich To: Thomas Duxbury Cc: sbnpsa@astro.umd.edu, Jian-Yang Li , "Edwin J. (GSFC-6901) Grayzeck" , thomas.h.morgan@nasa.gov, "William (HQ-DG000) Knopf" , Peter Thomas , "Kenneth P (382E) Klaasen" , Ludmilla Kolokolova Subject: Re: Shape model consistency metrics Tom, Thanks for the explanation.  Peter can fill in, but my memory is that his shape models develop in an organic fashion around an initial estimate for the origin.  Moving the origin is a somewhat awkward operation that invalidates a lot of ancillary data (e.g. SP-Kernels), and since it is not important it is not often done (generating a good SPK is a pretty intensive process and invalidating it for no gain is not an attractive option).  Most shapes come from flybys, so the exact location of the center is meaningless practically.  Geodesy investigations start with the 3D surface mesh, assume homegeneity or some other density distribution to calculate a center of mass and principal moments, and move on from there (e.g. I assume you have seen this before:  http://www.cs.berkeley.edu/~jfc/mirtich/massProps.html). Another issue is that Peter's canonical format is the R,Lat,Lon (planetocentric) format, so moving the center involves interpolation of the Lat,Lon grid points on the surface, which moves the calculated center again and messes up the control points, so finding an origin to match the center of mass or figure becomes a never ending task, again for no practical gain. So I guess that is "6) Other" (or maybe 2, 3 or 5 for the R,Lat,Lon issues) and these explanations from the reviewed data set may need some tweaking: Note that the center of figure is not exactly coincident with the coordinate origin. This offset has not been corrected in the present model as further refinement of the relative viewing geometry of the two missions may affect the shape model. The shape is such that it is not well represented by a triaxial ellipsoid, and the center of figure is not exactly coincident with the origin of the coordinate system, an issue that should be corrected in future versions of the shape model. I believe there is also an explanation of how spin pole and PM are determined already in dataset.cat. Btw, the script I sent out does not work if the center chosen is positioned on the outside-side of the plane of any of the triangular plates i.e. for some centers in a non-convex shape.  It could be trivially made to work (remove the abs() call) for any origin, even one outside the body, if the ordering of each triangular plate's vertices is known to be always the same wrt inside/outside; IIRC, VRML requires that, so it may already be true; Tony probably knows. Also, I assume you have seen this before:  http://www.cs.berkeley.edu/~jfc/mirtich/massProps.html Best regards, -b On Wed, Mar 6, 2013 at 2:10 PM, Thomas Duxbury wrote: [...] From tduxbury@gmu.edu Wed Mar 6 20:22:13 2013 Date: Wed, 06 Mar 2013 17:22:11 -0800 From: Thomas Duxbury To: drbitboy@gmail.com Cc: sbnpsa@astro.umd.edu, Jian-Yang Li , "Edwin J. (GSFC-6901) Grayzeck" , thomas.h.morgan@nasa.gov, "William (HQ-DG000) Knopf" , Peter Thomas , "Kenneth P (382E) Klaasen" , Ludmilla Kolokolova Subject: Re: Shape model consistency metrics Brian, like I said, most shape file descriptions had almost all of the descriptive information that I was ranting about and as a guess I mentioned that current shape models are an improvement to previous shape models that, as you state, other data sets are now tied to. There is no easy solution to what I propose and there is little to no impact if such a change to try to explain or recenter / re-orient the model. Therefore I stand by my formal PDS-SBN review board position: the model and documents are OK as is, no changes required and they should be certified for archive in the PDS SBN. If someone wants to add more descriptive information to the documentation, this is also OK by me. Tom