DG High definition data

28 views
Skip to first unread message

jai singla

unread,
Nov 27, 2024, 4:21:18 AM11/27/24
to Ames Stereo Pipeline Support
Dear All,

I have generated DEM at 15cm  spacing using worldview-3 stereo data with RPC inputs.

I utilized stereo command with RPC option with alignment method as none.

Generated DEM with data  is quite accurate.

I got valid pixel of 52.3% and there are gaps/holes in the DEM. (attached screenshot)

Can you pls suggest me to make a better dem.
image001.png

David Shean

unread,
Nov 27, 2024, 12:11:52 PM11/27/24
to jai singla, Ames Stereo Pipeline Support
Hi Jai,
The “native” GSD of the HD images is 15 cm at nadir, but I expect off-nadir will be worse, unless Maxar is running the super-resolution models for fixed a output GSD of 15 cm, regardless of the actual native WV-3 image GSD (>40 cm for larger off-nadir angles).  

There is a fundamental limitation with output point density in the triangulated cloud - matching is never perfect and you will not have a triangulated point every 15 cm, so it is not surprising that there are gaps in your DEM if you are attempting to post the output DEM at 15 cm.

I recommend you post your output DEM at 2-3x the “native” image GSD, so something like 30-50 cm at a minimum.  You can also explore the --search-radius-factor option in point2dem.
-David



--
David Shean
Civil and Environmental Engineering
University of Washington
https://www.ce.washington.edu/facultyfinder/david-shean

201 More Hall, Box 352700
3760 E. Stevens Way NE
Seattle, WA 98195-2700
Office: (206) 543-3105, Wilcox Hall 265
Pronouns: he, him, his

On Nov 27, 2024, at 4:21 AM, jai singla <jais...@gmail.com> wrote:

This Message Is From an Untrusted Sender
You have not previously corresponded with this sender.
See https://itconnect.uw.edu/email-tags for additional information. Please contact the UW-IT Service Center, he...@uw.edu 206.221.5000, for assistance.
<image001.png>

-- 
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/5259677b-acea-42b7-a1c2-d719d7bd907an%40googlegroups.com.
<image001.png>

jai singla

unread,
Nov 27, 2024, 11:45:05 PM11/27/24
to David Shean, Ames Stereo Pipeline Support
Thanks Mr David.

Let me try with new options and post the results. 
--
Regards,
Jai Singla

jai singla

unread,
Nov 28, 2024, 3:49:16 AM11/28/24
to David Shean, Ames Stereo Pipeline Support
Dear all,

By putting sampling at 3x and using search radius factor , DEM is generated with 82% valid pixels now.

I did not mention earlier but triangulation error cutoff is ~3.5 meter. How can I improve this?



--
Regards,
Jai Singla

Alexandrov, Oleg (ARC-TI)[KBR Wyle Services, LLC]

unread,
Nov 28, 2024, 11:47:02 AM11/28/24
to jai singla, David Shean, Ames Stereo Pipeline Support
>By putting sampling at 3x and using search radius factor , DEM is generated with 82% valid pixels now.

Glad there is an improvement.

>I did not mention earlier but triangulation error cutoff is ~3.5 meter. How can I improve this?

The triangulation error shows how far the rays are when they are supposed to intersect at a point. It is suggested to read this section https://stereopipeline.readthedocs.io/en/latest/tools/point2dem.html#orthoimage-and-error-image

Then you can try to produce the error image and colorize and plot it. Hopefully it will show no gradient across image, or else you need bundle adjustment.

If you want to reduce the triangulation cutoff, which will eliminate points in the DEM with higher such error, you can use the point2dem option:  --max-valid-triangulation-error.


From: ames-stereo-pi...@googlegroups.com <ames-stereo-pi...@googlegroups.com> on behalf of jai singla <jais...@gmail.com>
Sent: Thursday, November 28, 2024 12:49 AM
To: David Shean <dsh...@uw.edu>
Cc: Ames Stereo Pipeline Support <ames-stereo-pi...@googlegroups.com>
Subject: [EXTERNAL] [BULK] Re: DG High definition data
 
CAUTION: This email originated from outside of NASA.  Please take care when clicking links or opening attachments.  Use the "Report Message" button to report suspicious messages to the NASA SOC.



jai singla

unread,
Dec 5, 2024, 12:05:22 AM12/5/24
to Alexandrov, Oleg (ARC-TI)[KBR Wyle Services, LLC], David Shean, Ames Stereo Pipeline Support

Hi All,


I have following queries:


1.       As input data is 15CM DG stereo data , I want to generate very crisp building level DEM (Currently some of the buildings / vegetation is merged as one feature). Trying SGM matching for this. Can you suggest some better way? Can I use point2las??

2.       Currently, DEM is generated without using any input DEM. Still, getting good absolute accuracy . Shall I use input DEM for best results?


--
Regards,
Jai Singla

jai singla

unread,
Dec 12, 2024, 12:03:18 AM12/12/24
to Alexandrov, Oleg (ARC-TI)[KBR Wyle Services, LLC], David Shean, Ames Stereo Pipeline Support
Dear All,

Request your reply ....
--
Regards,
Jai Singla

Oleg Alexandrov

unread,
Dec 12, 2024, 11:38:46 AM12/12/24
to jai singla, Alexandrov, Oleg (ARC-TI)[KBR Wyle Services, LLC], David Shean, Ames Stereo Pipeline Support

1.       As input data is 15CM DG stereo data , I want to generate very crisp building level DEM (Currently some of the buildings / vegetation is merged as one feature). Trying SGM matching for this. Can you suggest some better way? Can I use point2las??


The DEM resolution cannot be as good as input image resolution. You can try --stereo-algorithm asp_mgm and --corr-kernel 5 5 or maybe 3 3 and see how it goes. 
 

2.       Currently, DEM is generated without using any input DEM. Still, getting good absolute accuracy . Shall I use input DEM for best results?



You can try, by cropping a small DEM clip mapprojeting on it, and see how it goes.

jai singla

unread,
Dec 12, 2024, 11:39:21 PM12/12/24
to Oleg Alexandrov, Alexandrov, Oleg (ARC-TI)[KBR Wyle Services, LLC], David Shean, Ames Stereo Pipeline Support
Dear All,

When I used parallel_stereo with asp_mgm, it generated plenty of tiles. it generated out-PC_sub.tif over a small region .
How can I generate DEM out of these tiles ?
--
Regards,
Jai Singla

Oleg Alexandrov

unread,
Dec 13, 2024, 1:14:58 AM12/13/24
to jai singla, Alexandrov, Oleg (ARC-TI)[KBR Wyle Services, LLC], David Shean, Ames Stereo Pipeline Support
It should produce an out-PC.tif at the end. Then you need to use point2dem as in the tutorial.

If you did not get out-PC.tif, that means it may have failed somewhere. 
Reply all
Reply to author
Forward
Message has been deleted
0 new messages