Georeference an image in 3D (VCS)

[PeterGebhardt]

Hi Everyone, we work in archaeology and are looking into a workaround to georeference stratigraphic profiles. We create large profiles using photogrammetry and export them. Now we try to georeference them in 3D, in a local scene using X,Y,Z. But even with an Vertical Coordinate System (VCS) activated the georeferencing option in the raster toolbox is greyed out. Anyone has a workaround? I mean its just referencing a picture in 3D, this should be possible right? Any help welcome 🙂

Cheers

Peter

[TravisSaladino]

From what I can tell looking at our help on georeferencing, it's only available in a 2D map. See https://pro.arcgis.com/en/pro-app/latest/help/data/imagery/overview-of-georeferencing.htm . I assume all the math is done in a planar coordinate system. 

Is there an issue you are seeing with the results of 2D georeferencing? Or are you hoping to get some other visualization output in 3D? 

Once an image is georeferenced, it can be draped over a 3D surface. And there might be other options too, depending on the data you have available. 

[BobBooth1]

I was thinking that a line could be digitized along the top of the profile and it could be extruded between the relevant elevations of the bottom and top of the profile image, and then the image could be texture mapped onto it. I expect that would take a custom tool.

@PeterGebhardt  - could you share a sample of the data?

[PeterGebhardt]

@TravisSaladino / @BobBooth1 

Thank you both for your answers. It is vertically oriented, we have no issues with Ortho and georeferencing planar top-down photos.
Here is a link to my G-Drive, it includes a GeoTIFF and DSM/DTM in TIFF as well. In this case it is a inner castle wall we'd like to have georeferenced and then printed in 1:20 so we can draw and create damage mapping.

https://drive.google.com/drive/folders/1w05rX5k-7r2TqkT6RQ6WQ537dmG4axEu?usp=drive_link

[BobBooth1]

OK, I've got those.

Do you have any contextual vector data? It would be helpful to have a plan view of the excavation area, or even a line feature that represents the section of castle wall. Alternatively, what are the horizontal and vertical dimensions of the image?

The two other TIFFs represent the topography of the wall surface?

How are you assembling the mosaic? 

Thanks!

[PeterGebhardt]

Hi Bob, thanks for your reply. I added a csv file with four GCP coordinates in it. GCP1 top left corner, GCP2 top right corner, GCP3 bottom right corner, GCP4 bottom left corner.  EPSG 2056 CH1903+/LV95

Location is Burgruine Freienstei: https://maps.app.goo.gl/h4jdYQQtipgrXFMF8

We use 3DF Zephyr to create the Orthos. You can ignore the DSM/DTMs - most important for us is to have the "profile" for documentation/drawing.

[BobBooth1]

Hi Peter,

I was able to make four points from your CSV using the XY Table to Point tool

https://pro.arcgis.com/en/pro-app/latest/tool-reference/data-management/xy-table-to-point.htm

with the optional Z field. When I tried the EPSG code you suggested, the result points fell in the ocean, so there is some issue there (could be some error on my part there). I did get four 3D points, which I added to a scene.

In my geodatabase, I created a multipatch feature class with a matching coordinate system.

I started an edit session on it, and used the editing tools in the scene to digitize a quadrilateral with vertices at each point.

With the multipatch feature selected, I used the Multipatch texture tool

to define a texture on the multipatch, loading the image and using the pan/rotate/zoom controls below the image preview to adjust the fit approximately.

This worked, but it was not very precise.

I got a resulting multipatch with the image on it.

Perhaps because I do not have a local DEM, or possibly due to my problem with the coordinate system, I'm having some trouble navigating the scene.

@PeterKnoop's suggestion of exporting a GLB object with texture looks like it would be a better approach.

[TravisSaladino]

@PeterGebhardt To better understand the data. and needs.. what is the orientation of the image? Is it an orthographic (top down photo) like a satellite image or is it a vertically oriented image like  a picture of a house looking at the side. More of what Bob is thinking about?  When you reference a stratigraphic profile... that would imply a vertically orientation to me. 

If that vertical orientation is the case, georeferencing is the not the correct tool to use here. Georeferencing is most applicable to orthographic images... or CAD like data.  And if that is the case I think Bob's approach is a good place to start. 

[PeterKnoop]

@PeterGebhardt during your photogrammetry step, are you able to output your desired content as a textured 3D object (e.g., a .glb, .gltf, .fbx, .obj), or other file format supported by ArcGIS? 

If so, you could place it as a 3D object in a Local scene in Pro, alongside your other GIS data. You can then use the regular feature editing tools (i.e., move, scale, rotate) to position it correctly in the real-world, relative to your other data, such as your control points, essentially 3D georeferencing it.

[PeterGebhardt]

Hi Peter, thank you for your input. I was thinking about the same. But if I use the whole model with high res textures as we need them, the file size would explode. But I'll try remove anything of the model except the single walls and just export those parts as low polygon OBJ since OBJ should even be georeferenced according to 3DF Zephyr.