point2dem failure

[Catherine Hanagan]

Hi! 

I have been working with WorldView images on an HPC (and my experience with an HPC is limited. I've run the following commands to mapproject and then run parallel_stereo, but when I run point2dem I receive this message several times until point2dem exists:

"DEM extent estimation failed. Check if your cloud is valid. Trying again with finer sampling."

Any suggestions are greatly appreciated. Here are the commands preceding point2dem:

dem_geoid --geoid egm2008 --reverse-adjustment output_hh.tif -o dem

proj="+proj=utm +zone=45 +datum=WGS84 +units=m +no_defs"

mapproject -t rpc --threads 60 \

--tr 0.5 \

--t_srs "$proj" \

dem-adj.tif ./20240116stereo/24JAN16050550-P1BS-050278578010_01_P002.TIF ./20240116stereo/24JAN16050550-P1BS-050278578010_01_P002.XML ./20240116stereo/24JAN16050550-P1BS-050278578010_01_P002.map.TIF

mapproject -t rpc --threads 60 \

--tr 0.5 \

--t_srs "$proj" \

dem-adj.tif ./20240116stereo/24JAN16050706-P1BS-050278578010_01_P001.TIF ./20240116stereo/24JAN16050706-P1BS-050278578010_01_P001.XML ./20240116stereo/24JAN16050706-P1BS-050278578010_01_P001.map.TIF

parallel_stereo -t dg --processes 100 --corr-memory-limit-mb 256 --cache-size-mb 2048 \

--alignment-method none \

--stereo-algorithm asp_mgm \

--subpixel-mode 3 ./20240116stereo/24JAN16050550-P1BS-050278578010_01_P002.map.TIF \

./20240116stereo/24JAN16050706-P1BS-050278578010_01_P001.map.TIF \

./20240116stereo/24JAN16050550-P1BS-050278578010_01_P002.XML \

./20240116stereo/24JAN16050706-P1BS-050278578010_01_P001.XML \

run_post/run dem-adj.tif

proj="+proj=utm +zone=45 +datum=WGS84 +units=m +no_defs"

point2dem --t_srs "$proj" --tr 0.5 --dem-hole-fill-len 100 --orthoimage-hole-fill-len 100 run_post/run-PC.tif --orthoimage run_post/run-L.tif

[Oleg Alexandrov]

Your command is correct.

This is a symptom of no points in the cloud being created, and point2dem fails on the empty cloud.

My best guess is that instead of using a stereo pair, you are using two portions of the same observation (parts P001 and P002). These are meant to either be merged with dg_mosaic into a single left image, and then a second unrelated right image is manufactured the same way, or the P001 portion from here is run together with the P001 portion from the unrelated image.

To check if this is the case, it is suggested to open your left mapprojected image, right mapprojected image, from above, and the DEM you mapproject onto, in stereo_gui, side by side. They can also be overlaid on top of each other with georeference info from the View menu. 

You should either see the second image continuing the first one, which is what I expect, or all 3 appear in the same place and are similar enough, which would be a good outcome when stereo would succeed.

--
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/0259e420-5344-4ff5-bab0-265666038d4en%40googlegroups.com.

[Oleg Alexandrov]

Looking at this more carefully, there seems to exist some dissimilarity in the first part of image names, while the second part is the same. So, perhaps, instead of first and second part of the same image, what is used here is the first part of one image and 2nd part of second image.

In either case, a visual inspection would confirm what is going on. The expectation of stereo is that the left and right image observe the same site, which is also what the input DEM is for, and these are from different perspectives.