Purpose of Test: Take a single maplet in center of a tiled area and perform many calculations of "FindHeights" on that maplet using various options within LITHOS.
Image suite: Same as OptimalF2.
Results: The solution diverged away from truth when performing the basic "FindHeights" (i.e. options 8 and 2) whether or not we aligned (101n0y) after each calculation or updated the maplet height (v1). Using stereo (options 8,2,6) drifted away a bit but then oscillated. The neighboring maps were never altered, and including them (options 8,2,1) really nailed down the solution. Including neighboring maps as well as stereo (options 8,2,6,1) created a bit more drift and oscillation than just with neighboring maps, but the solution was still pretty good. Including extended shadow masking (options 8,2,6,1,7) gave a better solution than when it wasn't included. No limbs were found (option 4), so it was not tested.
See Keynote presentation in SPC-ORex/Test_Info_documents/Test06/ for more details.
Conclusions: Using the basic options (8 and 2) created a tilt that couldn't be corrected. One or two calculations aren't too bad since it takes several calculations to drift a significant amount, but don't get carried away. Stereo improves the solution a significant amount, and extending the shadows improves the solution a minor amount. In this test including neighboring maplets improved the solution the most, but partly because the surrounding maplets were not changing. During a normal iteration the neighboring maplets can also drift, so in a normal situation we may not see as much of an improvement from including neighboring maplets.
The biggest conclusion is that apparently including the stereo option is required to keep the solution from drifting away from the truth. Including neighboring maplets helps as well, but this test was not able to quantify the improvement.
Looking into tilt
Heat map showing the presence of tilt. The tilt is not about the center, but on the right. This was created from test4, which included differential stereo and photoclinometry, but nothing else. The colormap is in meters.