Forum

Author Topic: Align new cameras into existing model without creating new tie points  (Read 2587 times)

rdunner

  • Newbie
  • *
  • Posts: 2
    • View Profile
I am interested in using Metashape to determine the position of a camera with respect to a terrain which I have already reconstructed using a previous set of pictures. My goal is to determine the position of a drone over a landscape with high accuracy using 4k video, so I am not interested in improving the model with the new pictures, but just want to find out the new camera coordinates and angles with high accuracy. I have been able to successfully align the new pictures by keeping the old key points, but the process is exceedingly slow for my application. I suspect that most of the time is taken by the optimization of the new tie points. Moreover, the new pictures are taken from basically the same point, which apparently gives a hard time to the alignment algorithm. I am using coded targets to improve the alignment, but the speed is still slow.

Is there a way to speed up the alignment process by just matching points and finding the camera locations while skipping other tasks like determining new tie points or optimizing the camera calibration?

I am using Metashape-pro 1.5.3

Thanks!

Rolando

Kiesel

  • Sr. Member
  • ****
  • Posts: 332
    • View Profile
Re: Align new cameras into existing model without creating new tie points
« Reply #1 on: October 17, 2019, 08:13:28 AM »
Hi Rolando,

you could try to align your new photos by coded markers alone (Tools/Markers/Detect Markers) if you still have them.


Regards,

Kiesel

rdunner

  • Newbie
  • *
  • Posts: 2
    • View Profile
Re: Align new cameras into existing model without creating new tie points
« Reply #2 on: October 17, 2019, 06:06:40 PM »
Thanks Kiesel.

I do have the coded targets detected in all my frames, but I don't know how to align them using only the markers. How can I do that?

Regards,

Rolando

Kiesel

  • Sr. Member
  • ****
  • Posts: 332
    • View Profile
Re: Align new cameras into existing model without creating new tie points
« Reply #3 on: October 18, 2019, 11:33:28 AM »
Hi Rolando,

detected coded markers are added as valid tie points. So they are used in normal alignment procedure.  (Unfortunately, I'm not in my office to make a little test for you. You have to wait until Monday if needed).
Do you know this tutorial?
https://www.agisoft.com/pdf/PS_1.1_Tutorial%20(IL)%20-%20Coded%20Targes%20and%20Scale%20Bars.pdf



Regards,

Kiesel

Kiesel

  • Sr. Member
  • ****
  • Posts: 332
    • View Profile
Re: Align new cameras into existing model without creating new tie points
« Reply #4 on: October 22, 2019, 09:17:25 AM »
Hi Rolando,

you can try the following:
1. put your new photos in a new folder
2. add a new chunk in your project and add new photos by choosing the folder
3. detect coded markers in this new chunk (Tools/Markers/Detect Markers...)
4. load marker coordinates.
5. align cameras in these new chunk.
6. align chunks (Workflow/Align chunks) method: Marker based
7. merge chunks (Workflow/Merge Chunks). This will generate a new merged chunk.

Regards,

Kiesel

EDIT: but something looks wrong, double coded markers with same name at different locations, hmm.
« Last Edit: October 23, 2019, 09:42:55 AM by Kiesel »