Pleiades RPC stereo invalid elevations

[Sotiris Valkaniotis]

Hello,

as a begginer in AmesStereo, I did a trial run to test the stereo satellite dsm extraction using a stereo-pair of Pleiades, based on the command line suggestions from the manual/book.

First run using stereo -t rpc was completed (IMG_.tif and RP_.xml files as input), matching was along the expected levels for the landscape conditions, but elevations of the output-DEM.tif were way off [-2000 to -4500m] due to an apparent left-to-right ramp (disparity-epipolar error?).

Second run I tried mapproject to anchor in real elevation (using the ALOS DSM tif tile that contains the AOI of the stereo-pair), but the resulted DSM was similar with the previous run.

Apart from the expected ./stereo.default error message, there wasn't any apparent problems during execution.

Here are my commands for mapproject/stereo rpc:

• mapproject -t rpc --t_srs "+proj=utm +zone=31 +south +units=m +datum=WGS84 +no_defs" N045E005_AVE_DSM.tif IMG_PHR1B_P_201308311041306_SEN_665631101-001_R1C1.tif RPC_PHR1B_P_201308311041306_SEN_665631101-001.XML left-mapped.tif

• mapproject -t rpc --t_srs "+proj=utm +zone=31 +south +units=m +datum=WGS84 +no_defs" N045E005_AVE_DSM.tif IMG_PHR1B_P_201308121038000_SEN_665610101-001_R1C1.tif RPC_PHR1B_P_201308121038000_SEN_665610101-001.XML right-mapped.tif

• stereo -t rpcmaprpc --alignment-method none left-mapped.tif right-mapped.tif RPC_PHR1B_P_201308311041306_SEN_665631101-001.XML RPC_PHR1B_P_201308121038000_SEN_665610101-001.XML pleiades/pleiades N045E005_AVE_DSM.tif

• point2dem pleiades/pleiades-PC.tif

.

Is there a step or a parameter I am missing?

RPC files are native ones delivered with the images, images were converted to TIF from native JP2 files.

[Alexandrov, Oleg (ARC-TI)[SGT, INC]]

Howdy,

The position of the output stereo really depends on the quality of the inputs. On its own, the -4500 elevation shift is not terrible, you can recover from it using our pc_align. Or, you can do bundle_adjustment while collecting GCP, per our doc, but that would be more manual work. The pc_align tool should really do the job.

I also see noise and missing areas in your DEM.  You can later try --stereo-algorithm 2 or --subpixel-mode 3, but you can do these first on a small area (for example, chosen in the GUI) as these may run slow. 

Oleg

---

From: 'Sotiris Valkaniotis' via Ames Stereo Pipeline Support [ames-stereo-pi...@googlegroups.com]
Sent: Wednesday, February 20, 2019 3:59 AM
To: Ames Stereo Pipeline Support
Subject: Pleiades RPC stereo invalid elevations

--
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 on the web visit https://groups.google.com/d/msgid/ames-stereo-pipeline-support/548c719a-a999-4e97-ad12-aee0a34726b7%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Sotiris Valkaniotis]

Thanks for the quick reply.

If there isn't any omission or error from my execution of stereo, I'd be actually dissapointed by the results, as this is a native stereo high-quality pair of images. Same pair gave an excellent DSM using other software (PCI & MicMac), w/only the RPC model.

I'd expected that using the reference DEM, it would anchor the model close to real elevation values - or it only assists the matching process?

Sotiris Valkaniotis

Τη Τετάρτη, 20 Φεβρουαρίου 2019 - 7:12:36 μ.μ. UTC+2, ο χρήστης Alexandrov, Oleg (ARC-TI)[SGT, INC] έγραψε:

Howdy,

The position of the output stereo really depends on the quality of the inputs. On its own, the -4500 elevation shift is not terrible, you can recover from it using our pc_align. Or, you can do bundle_adjustment while collecting GCP, per our doc, but that would be more manual work. The pc_align tool should really do the job.

I also see noise and missing areas in your DEM.  You can later try --stereo-algorithm 2 or --subpixel-mode 3, but you can do these first on a small area (for example, chosen in the GUI) as these may run slow. 

Oleg

From: 'Sotiris Valkaniotis' via Ames Stereo Pipeline Support

Sent: Wednesday, February 20, 2019 3:59 AM
To: Ames Stereo Pipeline Support
Subject: Pleiades RPC stereo invalid elevations

Hello,

as a begginer in AmesStereo, I did a trial run to test the stereo satellite dsm extraction using a stereo-pair of Pleiades, based on the command line suggestions from the manual/book.

First run using stereo -t rpc was completed (IMG_.tif and RP_.xml files as input), matching was along the expected levels for the landscape conditions, but elevations of the output-DEM.tif were way off [-2000 to -4500m] due to an apparent left-to-right ramp (disparity-epipolar error?).

Second run I tried mapproject to anchor in real elevation (using the ALOS DSM tif tile that contains the AOI of the stereo-pair), but the resulted DSM was similar with the previous run.

Apart from the expected ./stereo.default error message, there wasn't any apparent problems during execution.

Here are my commands for mapproject/stereo rpc:

• mapproject -t rpc --t_srs "+proj=utm +zone=31 +south +units=m +datum=WGS84 +no_defs" N045E005_AVE_DSM.tif IMG_PHR1B_P_201308311041306_SEN_665631101-001_R1C1.tif RPC_PHR1B_P_201308311041306_SEN_665631101-001.XML left-mapped.tif

• mapproject -t rpc --t_srs "+proj=utm +zone=31 +south +units=m +datum=WGS84 +no_defs" N045E005_AVE_DSM.tif IMG_PHR1B_P_201308121038000_SEN_665610101-001_R1C1.tif RPC_PHR1B_P_201308121038000_SEN_665610101-001.XML right-mapped.tif

• stereo -t rpcmaprpc --alignment-method none left-mapped.tif right-mapped.tif RPC_PHR1B_P_201308311041306_SEN_665631101-001.XML RPC_PHR1B_P_201308121038000_SEN_665610101-001.XML pleiades/pleiades N045E005_AVE_DSM.tif

• point2dem pleiades/pleiades-PC.tif

.

Is there a step or a parameter I am missing?

RPC files are native ones delivered with the images, images were converted to TIF from native JP2 files.

--
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.

[Alexandrov, Oleg (ARC-TI)[SGT, INC]]

Yes, the input DEM only assists with matching. I don't know why our software gives you a shifted DEM and MicMac does not. I've not heard from other folks using Pleiades images about such a shift. You can always try to recreate the rpc models using our cam2rpc, but I would not think the problem is there. And you can try the other options I suggested for improving the stereo quality. 

---

From: 'Sotiris Valkaniotis' via Ames Stereo Pipeline Support [ames-stereo-pi...@googlegroups.com]
Sent: Wednesday, February 20, 2019 9:42 AM

To: Ames Stereo Pipeline Support

Subject: Re: Pleiades RPC stereo invalid elevations

To unsubscribe from this group and stop receiving emails from it, send an email to ames-stereo-pipeline...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/ames-stereo-pipeline-support/dfab4a56-860c-4d86-857b-4783e955cf01%40googlegroups.com.

[jai singla]

Hi All, 

I also have Pleiades sample Tri-stereo data along with rpc info in WGS84 format.

stereo left.tif centre.tif right.tif does not result in to aligned and positive DEM.. I am also getting shifted and negative DEM.

Also tried mapprojecting option, results are better but again off and negative...  

 

I even passed ALOS data as refdem, marked match points  but still no luck...

I tried to align the data but it says "no point to minimize"

Awaiting further suggestions...

[David Shean]

Hi Sotiris and Jai,

Two suggestions:

1. I believe ALOS (AW3D30) provides elevation values as orthometric heights above the EGM96 geoid.  ASP can handle different vertical datums but I’m not sure this functionality is built into mapproject, and the RPC implementation is expecting heights relative to WGS84 ellipsoid.  

Using EGM96 heights can introduce significant horizontal offsets in the mapproject orthoimages depending on the geoid offset for your location.  One solution is to run `dem_geoid —reverse-adjustment refdem.tif` to obtain ellipsoidal heights, then rerun your mapproject with that refdem-adj.tif.  The other is to use a reference DEM already corrected relative to WGS84 ellipsoid, like some of those available from the OpenTopography Global DEM API: https://opentopography.org/developers

2. Check the product level of your Pléiades images.  I’ve had issues with Pleiades “Projected products” (equivalent of Maxar “ortho-ready” L2A images) instead of the expected “Primary products” (L1B images).  It seems less likely that this issue will produce the described offsets and “ramp” but worth checking.

Screenshots, logs and excerpts from the XML metadata always help troubleshoot.

-David

To view this discussion on the web visit https://groups.google.com/d/msgid/ames-stereo-pipeline-support/a74bd3cc-5c20-4e61-a06c-24a3c59f3adfn%40googlegroups.com.

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

Sorry for the pain.

> I tried to align the data but it says "no point to minimize"

This means that your --max-displacement value in pc_align is too small. When it tries to align two clouds, it needs to find matches from the first to the second.  This value tells it how far to look. 

If you have a elevation difference of say 2000 m, you can use --max-displacement 5000, for example. It also matters how big the horizontal shift is. 

You can experimenting with several values of --max-displacement and note also the messages pc_align prints. 

---

From: ames-stereo-pi...@googlegroups.com <ames-stereo-pi...@googlegroups.com> on behalf of jai singla <jais...@gmail.com>
Sent: Wednesday, March 10, 2021 4:07 AM
To: Ames Stereo Pipeline Support <ames-stereo-pi...@googlegroups.com>
Subject: [EXTERNAL] Re: Pleiades RPC stereo invalid elevations

 

To view this discussion on the web visit https://groups.google.com/d/msgid/ames-stereo-pipeline-support/a74bd3cc-5c20-4e61-a06c-24a3c59f3adfn%40googlegroups.com.

[jai singla]

Hi Shean,

1. ALOS (AW3D30) version we are using is containing WGS84 ellipsoidal heights.

2. You are right it is having processing level "Projected " Why it shall have problem and how did you solve this ?

--

Regards,
Jai Singla

[Etienne Berthier]

Jai,

Arghh this is unfortunate.

you could try to contact the Airbus customer service (through whic you ordered the data) and ask them to provide the PRIMARY Geometric processing level. This is the level you would like to use to generate DSM using ASP (or any other tools).

I hope they will reply positively.

To view this discussion on the web visit https://groups.google.com/d/msgid/ames-stereo-pipeline-support/CALB03WUnFykuKKj_LH_WS75oT%2BemX1FzqZ%3Dr%2BbR1AEdpiJEcKA%40mail.gmail.com.

[jai singla]

Hi All,

alignment offset  results of generated dsm vs image ..

I tried pc_align with max displacement of 1000. 

Attached are the results...

[jai singla]

Attached image

[jai singla]

Hello All,

Any updates on logs / screen shot..

What is the exact issue in projected data?

Do we need only Primary Geometric processing level product to proceed ?