No DSM/point cloud tiles results with Pleiades data

Question

No DSM/point cloud tiles results with Pleiades data

Thomas-Brd opened this issue 7 years ago · comments

Hi,

I am working on DSM generation from Pléiades data with s2p but I have some bad results that I do not understand.. Indeed, all the points clouds (and DSM) of each tile are not calculated. The following files are missing in the output files for "empty" tiles:

Height_map.tif
rectified_disp.tif
rectified_mask.png
rpc_err.tif

These files are missing in the "pair_2" folder of the affected tiles.

Is this problem related to the configuration of the starting parameters?

I have provided you as an attachment the initial configuration file (config.json) I used, the output configuration file (config_out.json) and the resulting DSM.

Thank you.

config_and-results.zip

Thomas.

gfacciol · Answer 1 · Tue Mar 13 2018 19:42:50 GMT+0800 (China Standard Time)

Dear Thomas,
The configuration seems good.
If the rectified_disp.tif files are missing only in the pair_2 directories,
then we might have a problem with the fusion.
To help you out, could you send us the content of one of the tile subdirectories with problems?

Have you tried running the pipeline independently on?

pair_1: Plei-001.tif -- Plei-002.tif
pair_2: Plei-001.tif -- Plei-003.tif

Please send us also the logs of all the tiles with:

zip logs.zip find -name *.log

Best,
Gabriele

Thomas Bernard · Answer 2 · Tue Mar 13 2018 21:26:33 GMT+0800 (China Standard Time)

Thank you for this quick answer. Here is a link to the tile subdirectories with problems:
https://drive.google.com/open?id=1ri1DuSIN_J8yDCaPSG1ymc5lBSjtWVEf

I did not try to run the pipeline independently yet but I will try to do it.

Here is the logs of the tiles:
zip_logs.zip

Best,
Thomas

gfacciol · Answer 3 · Tue Mar 13 2018 21:54:32 GMT+0800 (China Standard Time)

Thanks Thomas, I found the problem in the pair_2/stdout.log The stereo matching fails, probably due to an out of memory. RUN: mgm -r -41 -R 94 -s vfit -t census -O 8 /home/psteer/Desktop/s2p/testdata/input_triplet/whataroa/../../testoutput/output_triplet/whataroa/tiles/row_0033498_height_1000/col_0004622_width_1000/pair_2/rectified_ref.tif /home/psteer/Desktop/s2p/testdata/input_triplet/whataroa/../../testoutput/output_triplet/whataroa/tiles/row_0033498_height_1000/col_0004622_width_1000/pair_2/rectified_sec.tif /home/psteer/Desktop/s2p/testdata/input_triplet/whataroa/../../testoutput/output_triplet/whataroa/tiles/row_0033498_height_1000/col_0004622_width_1000/pair_2/rectified_disp.tif 01234567 01234567Killed The reason why this doesn't happen on the pair_1 is that the disparity range in that case is smaller. You have two options to fix this issue. 1) Quick&Dirty: recuce the size of the tiles: 700x700 instead of 1000x1000, should reduce in half your memory usage. "tile_size": 1000, 2) The memory usage of the MGM correlator was addressed in a recent pull request. https://github.com/gfacciol/s2p/tree/bump_mgm_multi You could fetch it and recompile mgm. To configure the new version of the correlator use: "matching_algorithm": "mgm_multi", Best, Gabriele

…

On Tue, Mar 13, 2018 at 2:26 PM, Thomas-Brd ***@***.***> wrote: Thank you for this quick answer. Here is a link to the tile subdirectories with problems: https://drive.google.com/open?id=1ri1DuSIN_J8yDCaPSG1ymc5lBSjtWVEf I did not try to run the pipeline independently yet but I will try to do it. Here is the logs of the tiles: zip_logs.zip <https://github.com/MISS3D/s2p/files/1807005/zip_logs.zip> Best, Thomas — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#148 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AExUamtD0wqsLSFzFCY4vlvJh6spGA0jks5td8kKgaJpZM4SofhW> .

Thomas Bernard · Answer 4 · Tue Mar 13 2018 22:45:56 GMT+0800 (China Standard Time)

Ok thanks! I will try the "mgm_multi" option.

All the best,

Thomas