Forum

Author Topic: Using Spherical Panoramas /equirectangular images as an input for 3D reconstruct  (Read 12561 times)

bisenberger

  • Sr. Member
  • ****
  • Posts: 329
    • View Profile
    • Digital Mapping & Graphics
Hello, xabierr.
Here is how Civetta works, with a click the camera will automatically turn 360 degrees can capture the scene, we recommend our users to capture measurement pairs( it means you will capture 2 panoramas with the exact same location but different height). The distance between each spot should be approximate 5-10 meters. In that case, it will automatically make sure it has enough overlaps.

The idea of using panorama in Agisoft is because it ensures a very high level of overlapping and it is a standardised process, so our users do not need to worry about how much overlap they will need to have to ensure the quality of the models. In addition:
1. It will be easier for the computer to recognise the common area between two different panoramas because they have a higher level of overlaps in between.
2. It means for measurement pairs, they only need to calculate one vector for two. So for 4 panorama images, Agisoft only needs to recognise 2 vectors. It is completely different from using your camera and captures thousands of photos, with thousands of vectors needed for the calculation.

In short, it ensures at least 60% overlapping.

Christiane Zhao
Feel free to follow me in LinkedIn: https://www.linkedin.com/in/christiane-zhao-a073b4130/
Hi Christiane,
Interesting thread  8). If I understand correctly you take two panoramas at the same X, Y (horizontal) location and vary the z (vertical) location by 5 to 10 meters. That is a pretty good distance to raise the camera on a stable platform, how is this accomplished?
Bill
Digital Mapping & Graphics LLC
https://digital-mapping.net/

Shadow

  • Newbie
  • *
  • Posts: 24
    • View Profile
    • My 3D scans on Sketchfab
Hello Bill,
If I understand correctly you take two panoramas at the same X, Y (horizontal) location and vary the z (vertical) location by 5 to 10 meters. That is a pretty good distance to raise the camera on a stable platform, how is this accomplished?
Bill
Note exactly. Check out Christiane's Youtube channel, there are a couple of making-of videos :
https://www.youtube.com/channel/UCAIY9rkqm8viTedXDvmLUWw/videos
Basically :
- you take two panoramas at the same X, Y, the Z only varies by around 1 meter (depending on your tripod) ;
- you then move the camera to another location by around 5 to 10 meters in the X/Y plane.