Height Control Points

[RainerBrechtken]

Is there a possibilty to introduce height control points into a Photoscan project?
Thanx a lot and best regards
Rainer

[Alexey Pasumansky]

Hello Rainer,

Currently it is only possible to use control points with all three coordinates. Height or planar control points are not supported.

[RainerBrechtken]

Dear Alexey,
that's a pitty, especially for creating precise elevation modells...
But thank you for your promt response.
Best regards
Rainer

[Mohabon]

+1 by me.

Height control point would help a lot in some particular conditions.

Example:
- i recently made an UAV survey of an area 2km by 2km. Placing GCPs was ok for central and southern area, while northern area was nearly impossible because of swamps.
- At the end of PS processing, i have a DEM + Ortomosaics which is quite accurate for XY positioning in north, central and south area
- BUT The DEM elevation is accurate only on central and south area, while north is very crap, which is perfectly understandable.
- Using a big wood tower + theodolite i estimated some AUXILIARY GCPs in the unaccesible area.

Since XY accuracy eas already good, it was difficult for me to import these new AUXILIARY points into PS to improve DEM elevation accuracy without reducing the already good XY accuracy.

So "Z-only GCPs" would be a  excellent improvement in my opinion!

[Porly]

Hello,

Optimize your project with all available gcp to reach a good xy quality.

Measure a point in the locality (where you want to add height control) in picture in at least 2 photos.

Export or note the estimated coordinates of these points.

XY should be good, just replace the height with the altitude you have for your height control points.

Load the project before the optimizing step.

Optimize the project with measured xy (from Photos) and the edited z values.

That should have the same effect since the xy coordinates have been estimated after the first optimizing.

The weight of xy in this case should not have influence on the result of the bundle block adjustment.

I have tested it, works fine.

BR

Paul

[Mohabon]

hello and thanks for your support, but your workflow is unclear to me.

Measure a point in the locality (where you want to add height control) in picture in at least 2 photos.

When you say "measure", how do could i do? At the end of alignment and first optimization, i have only a Sparse Cloud, so i dont know how to measure a target.
Is there an hidden command i dont know?

Thanks and regards,

MoH

[Porly]

Hello,

even when you have only the sparse point cloud you can set markers. If you have markers in at least two photos switch to "view estimated"-view in ground control pane. There you will see estimated coordintaes for your height control points (xyz).

But you have to make it twice to reach a good xy quality.

optimize the project with all gcp (xyz) you have. Measure (set markers) on position where you want to use height control points.

Load project before the first optimizing step and import the coordinates you have marked with the edited height for your height control points.

I hope it is clearer right now...

Best regards

Paul

[Mohabon]

Sorry i appreciate your help, so i dont want to look unpolite when i say it is not clear because maybe you jumps some steps in the workflow, assuming that i am more prepared than i am! 

1) I have my project: images imported, aligned and optimized with GCPs.
2) I save my project to have this status as Base
3) Since i try to work well, each point in the sparse point cloud has 7-10 reprojections, so it is available in 7-1o photos
4) Following your suggestion, i go in Model---Sparse Point Cloud---Right-Click--- Create Marker. Of course i dont see exactly where is placed this new GCP because sparse-point-cloud is low resolution. Or maybe should i create a marker inside a photo where i can exactly see where to place and find a suitable well known spot common to many photos?
5) Then i right-click on this new GCP i created at point 4 and i select "Filter photo by points" and i find only photos where this new GCP is available
6) I open each photo and i move this GCP to a point that i find well-known in all other photos (for example beside a tree). it takes quite long because each point does not move automatically on spot chosen. At the end of this phase, the estimated error is HUGE, dunno why.
7) Now i should export in tXT format all GCPS, so i have freshly new XYZ (real + height control points)
8 ) At this point, i am quite lost: accordingly with your explanation, i should load the previous project before the first optimization, which is not clear: do you mean project right after alignment?
9) no i edit the TXT file, dont touch XY and edit Z to the desired value.
10) reimport GCPs, optimize and things should be ok.

I think this workflow should work, but maybe there is something inbetween that makes it impossible for me.

Can you please clear it a bit further?

Thanks again and kind regards,

M

[Porly]

Hello Mohabon,

1)ok
2)ok
3)ok
4) I would "measure" these point directly in all available photos , and not in the model of the sparse point cloud (the xy coordinate should be fine this way).
5)ok
6)the error is big because the marker is checked, when marker without coordinates  is checked he gets the coordnates 0,0,0 (XYZ) in "view source"-view. Please uncheck this marker and you will see that the error will disappear.
7)export only the estimated coordinates of the new points!
8 ) yes, load the project before optimization, right after alignement.
9) now please pay attention. You cannot import all markers from the estimated view of the previous project (point 7), only these for the height points. The coordinates of GCP's you used in point 1) for optimization have to be the same.

Add only the new height points with the estimated XY and the edited Z to the Base Project (point 1)).

10) optimize, everything should be fine right now

Fell free to ask if there are more questions. I know that sometimes my explanations are not so easy to understand. It is not my biggest talent

Best regards

Paul

[Mohabon]

Hello again,

i start having a clue about your workflow.

1) I import photos, i do alignment. Then i save the project. This is Project-A.psz
2) I import real GCPs (point 1 to point 10 as example), i do optimization and i have well georeferenced XYZ model. I save this project and this is Project-B.psz
3) Since there is a doubt area in the model, then i go inside, i open some photos of this area and i find a well-suitable point which is common to 2+ photos. I place marker Point 11 in this spot. Then i place marker Point 11 in all photos available.
Note: if you dont check the new Point 11 in the Markers Panel, then its  ESTIMATED coordinate is not refined after you enable it on each photo. So it is needed to check-uncheck this point 11 for each photo.
At the end of this phase, I have point 1 to 10 and Point 11.
4) I export Point 11 estimated values, i edit Point 11 Z to the desired value. to obtain good solution
5) I open Project-A.psz, i import original markers and Point 11 edited marker, then i re-run optimization.

The game should be ok now. Confirmed?

Thanks and cheers!

[Porly]

Hi,

1) you can save Project-A after marking the real GCP's to not need mark them in the 2nd stage again.

3)you can "Update" transformation without checking point 11, so the refinement will go on. But check-uncheck should make the same.

That all sounds good, I think we got it

Best regards

Paul