point2dem offset using RPC model

22 views
Skip to first unread message

Haruki Hagiwara

unread,
Mar 28, 2025, 4:49:17 AMMar 28
to Ames Stereo Pipeline Support
Dear ASP community,

I'm trying to generate a DEM from ALOS/PRISM images using RPC model, and found x-y offset in the final DEM created by point2dem. Until parallel_stereo (i.e., -PC.tif), there seemed to be no big offset, so I guess point2dem causes the offset (x: ~300 pixels, y: ~60 pixels).
The original ALOS image was tiled into several images (IMG-0x.tif and IMG-0x_rpc.txt), so I merged them and embedded the RPC camera into the merged image (IMG.tif, and IMG_rpc.txt). When I generated a tiled DEM from tiled images, this shift doesn't happen in the final DEM.
Could you please give me some advice on how to fix this? I've attached here an image of the shift, comparing GoodPixel.tif and the final DEM.

The following are the methods to generate a DEM from merged ALOS/PRISM images.
---
bundle_adjust -t rpc $img1 $img2 -o $ba_out/out
parallel_stereo -t rpc --stereo-algorithm asp_bm --bundle-adjust-prefix $ba_out/out $img1 $img2 $im12 $refDEM
point2dem --t_srs "+proj=utm +zone=42 +ellps=WGS84 +datum=WGS84 +units=m +no_defs" $im12-PC.tif -o $im12-brend --dem-spacing 2.5
---
Best regards,
Haruki
Capture.PNG

Haruki Hagiwara

unread,
Mar 28, 2025, 5:03:24 AMMar 28
to Ames Stereo Pipeline Support
And the final DEM has ~600 m z-offset, compared with the DEM generated from tiled images.

2025年3月28日金曜日 9:49:17 UTC+1 Haruki Hagiwara:

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

unread,
Mar 28, 2025, 12:08:34 PMMar 28
to Haruki Hagiwara, Ames Stereo Pipeline Support
It is important to note that the PC.tif does not have honest georeference. It imitates the one from the input left image. 

The shift is likely due to inconsistency or inaccuracy in the input data. We have the pc_align program that can correct the shift.

We did not try that data type and not sure what its issues are. 

I also see some artifacts in your output DEM, which may be due to pre-processing, such as tiling, or mapprojection, etc. 

It is also suggested to try the asp_mgm algorithm.



From: ames-stereo-pi...@googlegroups.com <ames-stereo-pi...@googlegroups.com> on behalf of Haruki Hagiwara <kcblu...@gmail.com>
Sent: Friday, March 28, 2025 2:03 AM
To: Ames Stereo Pipeline Support <ames-stereo-pi...@googlegroups.com>
Subject: [EXTERNAL] [BULK] Re: point2dem offset using RPC model
 
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.
--
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/85638cb3-9a9b-4c36-b4e2-cacfcdc35485n%40googlegroups.com.

Haruki Hagiwara

unread,
Mar 30, 2025, 1:28:44 PMMar 30
to Ames Stereo Pipeline Support
Hi, thanks for the info.
After applying pc_align to the final DEM with proper --max-displacement value, the shift was solved.

Best,
Haruki

2025年3月28日金曜日 17:08:34 UTC+1 oleg.al...@nasa.gov:
Reply all
Reply to author
Forward
0 new messages