Processing Pleiades Level 1A

[lucia.and...@gmail.com]

Hello, 

I have ordered Pleiades Level 1A and I have been having some problems processing the data. 

I create georeferenced data from DIM.XML and apply atmospheric correction. Then I apply bundle_adjust with two stereo images and DIM.XML cameras (as suggested for Pleiades). However, the output of bundle adjustment is worse than the initial alignment. I have tries several runs with the range of setting but the output does not improve. Could this be caused by the input images? 

Any suggestions would be greatly appreciated. 

Many thanks, 

Lucia 

[Oleg Alexandrov]

I guess a first check should be if your input images and cameras are reasonably consistent. For that, one fetches a third-party DEM (https://stereopipeline.readthedocs.io/en/latest/next_steps.html#choice-of-initial-guess-terrain-model), converts it if needed to WGS84 (described in the same doc, a bit more down), then mapprojects the input raw images with DIM xml cameras onto the DEM (https://stereopipeline.readthedocs.io/en/latest/examples/pleiades.html#bundle-adjustment-and-stereo).

The left and right mapprojected images from one stereo pair should be in reasonable agreement as overlaid in a GIS tool. After that, one can try running stereo, with the original unprojected images, or with the mapprojected images, as described at that last link. This should be done without bundle adjustment for now, as normally Pleiades images are mostly already good even without that. 

If getting a good enough DEM, then one can attempt to bundle adjust as well, with the raw input (not mapprojected) images, then pass the option --bundle-adjust-prefix to the stereo command. If doing mapprojection, this option should be passed to the mapproject command too.

The hope is that this will show that things are good up to the bundle adjustment step. Normally that step should work too, I think. It is also suggested that one check that there's no mixup in inputs for this command, it should have left and right raw images and left and right cameras.

I think for now one should also avoid atmospheric correction. It is not that well tested for Pleiades.

Hopefully this will shed some light onto what is going on. 

--
You received this message because you are subscribed to the Google Groups "Ames Stereo Pipeline Support" group.
To unsubscribe from this group and stop receiving emails from it, send an email to ames-stereo-pipeline...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/ames-stereo-pipeline-support/aaf75f6f-fe3d-4406-93ec-825a22cd70e4n%40googlegroups.com.

[Oleg Alexandrov]

To add to previous email, the stereo DEM with or without bundle-adjusted images can still have some misalignment to a prior DEM, especially if the input DEM is not corrected to be relative to WGS84. The bundle adjust program only makes the cameras more self-consistent, not consistent with the ground. Our pc_align tool can fix any misalignment, which in your case should likely be at most 100 meters or so, I guess.