Difference between the Metadata (XMP) and Registration Files?

Development Asset Creation
,
1

What is the difference between the camera positions of the exported Metadata (.xmp) file and the Registration (.csv) file? Here are the outputs:

Metadata (.xmp) File

<xcr:Position>169.075617280029 -6.18064718007169 22.8503756266734</xcr:Position>

Registration (.csv) File

167.358408233295,-5.768472952332557,26.78456775075969

And which of these reflects the actual position of the cameras in the model?

Thank you in advance.

2

Hello Ellison,

is your project georeferenced? If not, then when exporting XYZ coordinates to csv via Registration, the coordinates are calculated with respect to the grid plane (for non-georeferenced project it is the only option). However XMPs are the coordinates from the center of the local Euclidean coordinate sytsem. 

3

Good day Zuzana,

I did input the supposed position of the cameras, but I think it is not enough, which is okay for my case. May I know where is the center of the local Euclidean Coordinate System, is it on the center of the model?

4

With the default position of the ground plane, the center of the local Euclidean coordinate system is placed in the center of the ground plane. You can check the position of the center of the coordinate system once you place the reconstruction region (via Reconstruction tab - Info Panel) to the center of the coordinate system (set x,y,z coordinates to 0) and defining the size of the region for example to 1 in each direction. The center of the cubic region will be in the center of the coordinate system. 

You can also define your own coordinate system by placing ground control points with specific coordinates (1 GCP for  the origin of CS, 2 GCPs to define the scale and 3 GCPs to define whole coordinate system).

 

5

Hi Zuzana, 

Would you be able to explain a a bit about how to define our own coordinate system by placing ground control points?

I found this video from RC team: https://youtu.be/5VyYLaNxHz0?t=131

At 2:21 in the video, they change from tie points to GCP and then they manually assign X, Y and Z values for all 4 points.

Assuming I have some measurements in the real world to use, how do I assign these these measurements into x, y, z values?

I have tried looking on these forums, but most mentions of GCPs use GPS coordinates or a theodolite to determine the X Y Z.

6

 

Hello Tobias,

I will try to explain it in more detail.

The coordinates used in the video are coordinates of the points in the created local coordinate system. You can create your own local coordinate system according to the distribution of your control points. The coordinate system is created in order to unify all models and place them to the same coordinate system.

In the video, the size of the rectangle defined by 4 vertices (4 control points) is 1.4 m x 1.0 m, therefore the coordinates are chosen to match the real-life measures. The origin of the local coordinate system was placed to point 1 (coordinates 0,0,0), X axis is heading to the point 2, therefore the X coordinate of that point is 1.4. Y axis is heading to the point 0, therefore the y coordinate of this point is 1.0. The coordinate system is defined as right-handed. The 4 points are in the XY plane therefore all Z coordinates are 0. 

If your points are distributed in the same way, you can simply change X and Y coordinates to match your measures. 

 

7

Hi Zuzana, Thank you very much for this great explanation. I have tested this by placing 3 control points. It seems to be creating a ghost camera in the middle of the air. 

And when I hit align, it will flip my model. and put those ghost cameras on the ground.

 

8

Parallel threads…

https://support.capturingreality.com/hc/en-us/community/posts/360012775591-Incorrect-alignment-with-camera-rig?page=1#community_comment_360001412812

 

I think that your points coordinates are bad. RC tries to solve these constraints and you end up with your model in a weird orientation.

 

9

And these are not “ghost cameras”, these are the control points you just added.

If you put the coordinates close to z=0, it is totally normal that they are on the ground. because z=0 is the ground…

 

10

These “ghost cameras”  seem to appear as soon as I change the control point from Tie Point to Ground Control Point. The coordinates that I set do not change the location of these “ghost cameras” 

11

The orange lines connect control points with the cameras(images) on which those are measured. Can you please check if the control points are correctly placed on all three images? Also, can you please share the GCP coordinates that you used?

12

I still cant figure out why there is 3 orange lines going into nowhere. There are no cameras there

I have checked the control points are they are placed correctly on all 3 images. 

there are the 3 points, using Local:1 Euclidean

  1. x = 0, y = 0, z = 0.835

  2. x = 0, y = 0, z = 0

  3. x = 0.835, y = 0, z = 0 

13

Try to reset your ground plane maybe you have a rotation applied here.

14

ok so coordinates for point 0, 1 and 2 in that order ?

this is a strange coincidence that you have the same distance (0.835) in x and z between your points. Are you sure of your measurements ? origin of your model will be point 1 (0 0 0), is this what you want ? from your coordinates, point 2 is right above the origin, so will define a vertical in your frame of reference. This doesn’t appear to be right based on your screenshot, this looks more like an horizontal line in your photo.

15

What is your 3 points left error after you update/realign ? (the orange number)

It should be 0 or very very close to 0. If not, something is wrong with your numbers. 

16

Tobias, would it be possible to share the project (project file and folder) and images with me so that I can examine it in more detail and find out what is causing this issue? If yes, please send me the download link via contact form on our webpage with the reference on this thread. Thanks

17

Zuzana I have uploaded the files to the contact form under “Other”. 

18

Hello Tobias,

thank you. I looked at your project. Those orange lines are residuals of the ground control points, not the lines connecting cameras with control points.

You have sent two sets of images, one folder contains also XMPs for the images. Have you loaded  the images to the RC along with those XMPs? If yes then the reason of such errors is that RC is using both GCPs and XMPs for georeferencing and if those are not consistent such errors appear. How were those XMPs created?

I tried to load the images without XMPs and align with the GCPs and the GCPs residuals were much smaller.


I also tried to load the images without XMPs to the new project, place GCPs and align, again project was georeferenced correctly. I can send you the project if you wish. 

19

Hi Zuzana thanks for checking. 

This is how I generated the XMP files: https://www.youtube.com/watch?v=-Yp6ZdCIE3k

I have an array of 18 cameras, and i am recording video on all 18 cameras, which are synced together perfectly. 

I believe that I need BOTH the xmps and the GCPs in order to get the alignment to be the same and to prevent any shift during export.

Is there a way to do this without both?

20

Hello Tobias,

If you wish to preserve the coordinate system of your rig then the workflow is to place GCPs to georeference your scene, then export XMPs and load them along with the images from the second dataset. You do not need to place the GCPs again in the second scene, this scene will be georeferenced using the position from the XMPs.