Printable Version of Topic

Posted by: Bjorn Jonsson Sep 4 2007, 08:12 PM

In time, just before the Cassini Iapetus flyby I did a preliminary DEM of a small area of Iapetus' surface:

The area it covers:



I created the DEM from a stereo pair using software I have been developing (see http://www.unmannedspaceflight.com/index.php?showtopic=4419 for more information).

The DEM is based on images with somewhat inaccurate viewing geometry - the pointing information in the PDS files really isn't accurate enough for my purposes. This is difficult to correct, in part because of Iapetus' somewhat irregular shape. I probably need to establish a global control point network to determine the exact pointing but there's a catch-22: To make it easier I really need accurate pointing information for at least one image but this is difficult to determine without a control point network...

In the DEM the difference between black (lowest) and white (highest) is 33 km. This is highly inaccurate because of the viewing geometry inaccuracies mentioned above. An obvious 'proof': Distances from Iapetus' center range from 785 to 818 km according to my stereo software. This cannot be correct, the best-fit triaxial ellipsoid has been determined to be 747.1 x 749.0 x 712.6 km. However, relative elevations over fairly small areas should be reasonably accurate so things like crater depths can be estimated. For example the depth of the crater centered at ~(380,220) is ~3.5 km.

A lot of additional imaging coverage is available so I should be able to make a bigger DEM once I have determined accurate pointing for the images I will be using.

Posted by: djellison Sep 4 2007, 08:32 PM

Beautiful work Bjorn!

Just for fun I used it as a bump map in Photoshop - I figure if it produces something that looks slightly realistic, it must be a good DEM.

Doug

Posted by: CAP-Team Sep 4 2007, 09:49 PM

Great work! How long does it take to calculate such a DEM?

Posted by: Bjorn Jonsson Sep 4 2007, 10:24 PM

This one took about 7 hours to compute on a 3 GHz machine. I will probably be making a somewhat higher quality version (less 'contoured' appearance) once I have more accurate camera pointing parameters. That one will probably take about 50 hours (!).

Posted by: Phil Stooke Sep 4 2007, 10:48 PM

Very nice work. Just goofing around - classes having not started yet - here's a composite of the DEM and Doug's relief version.

Phil

Posted by: Bjorn Jonsson Oct 9 2007, 08:56 PM

This is a significantly more accurate version of the DEM I posted at the start of the thread:

Altitudes are relative to a 748x713 km spheroid with black (0) representing -8.8 km and white (65535) 15.4 km. Compared to the original DEM this one is calculated to a higher subpixel accuracy and therefore looks less contoured. It took almost 70 hours (!!) to compute; seems I really need a supercomputer although I now have some ideas that may make this much faster. Also the viewing geometry is much more accurate than in the previous version but still a problem - I really need a 3D control point network.

The latitude range is -13.25° to 48.25° and the longitude range is 29.5° to 119.25°.

A color coded version:

Posted by: Phil Stooke Oct 9 2007, 09:33 PM

As before, a version of the above very nice DEM merged with shaded relief. I used the Photoshop emboss filter on the b/w DEM and merged it with the colour version.

Phil

Posted by: tedstryk Oct 10 2007, 12:38 AM

Wow! This is some great stuff.

Posted by: Bjorn Jonsson Dec 13 2007, 12:25 AM

A bigger and significantly improved version:

This is a mosaic of two DEMs computed from two stereo pairs. Altitudes range from -9.8 km (black) to 14.9 km (white) relative to a 748x713 km spheroid. Latitude ranges from -13.25 to 76.75 degrees and longitude from 29.5 to 144.125 degrees. The viewing geometry I used is probably somewhat more accurate than in the earlier versions and I am also in the process of establishing a rudimentary control point network. This DEM is computed using a new version of my software that is about 10 times faster than the earlier version so I was able to calculate it to a higher subpixel accuracy than earlier. Other improvements also result in slightly higher resolution. A consequence of these improvements is a bit more overall noise in the DEM. The image above is a slightly smoothed version.

A rendering generated using a color coded version:



An enormous impact basin measuring about 600 km across (~40% of Iapetus' radius) is obvious in the DEM/rendering. It is visible on close inspection in the original images I used. However, in the east this basin seems to have been destroyed. This is definitely a real feature in the DEM - there is an Icarus paper in press with DEMs that appear remarkably similar to mine (in fact they were highly useful as ground truth).