oleg-alexandrov
commented
2 years ago The recent consensus on this is that LRO MiniRF is not the most suitable sensor for stereo (and for the Moon LRO NAC and LOLA provide plenty of alternatives). In ASP the choice was made to approximate the rays as straight lines, at least for triangulation. Long-term, if desired to model convergence of curved rays, a reasonably simple solver can likely do the job.
The SAR CSM model produces curved rays. That may be physically more accurate, but stereo and bundle adjustment assume straight rays, so this is definitely a complication.
This also causes the CSM and ISIS SAR models to disagree as soon as one tests them at a height above the datum which is not 0.
This will need some careful studying. My plan is to first find a good SAR stereo pair, validate it with ISIS SAR cameras, and compare the produced DEM with one made perhaps with LRO NAC data. If those agree, and if the ISIS SAR model uses straight rays (which I think is true but only partially verified), maybe the solution is to make the CSM SAR model use straight rays too, hence sacrificing some of the physics in order to get a model usable in practice.