Axis Sign notation / model orientation ( Z-Axis) in Photoscan?

[ju523m]

Hi Photoscanners,

in bundle.out file from Photoscan  (Standard 0.8 B 877) there are positive Z-coordinates for points in front of the camera.

Bundler documentation states (please see http://phototour.cs.washington.edu/bundler/bundler-v0.4-manual.html):
"In the camera coordinate system, the positive z-axis points backwards, so the camera is looking down the negative z-axis, as in OpenGL."

So in a bundler conform bundle.out file there should be negative  Z coordinates for the 3 D Points. 
3D Models in Photoscan are first displayed "bottom-up" and this is the same with models exported to PLY or other format and then displayed in Meshlab for example. 

Rotating the model e.g 180 degrees around x-axis and then 180 degrees around z-axis makes it look bundler-like.

Please can someone definitely clarify axis sign notation / model orientation in Photoscan? Are the rotation parameters for the cameras (Photoscan structure or Omega-Phi-Kappa export of points) influenced by that or are they "bundler / OpneGL" conform ?

Many Thanks
Thomas

[ju523m]

Hi all,

no one any idea or hint on this? Can someone from Agisoft support team please give a statement concerning Z-axis notation / model orientation and the bundler.out files?

Many Thanks
Thomas

[Monserrat]

I think it must seek the support of Noah Snavely.  And not here  . Here you will find support for your tourism.

http://phototour.cs.washington.edu/

[Dmitry Semyonov]

Hello Thomas,

Point coordinates in bundler.out files are stored in the world coordinate system. It is different from the camera coordinate system (actually each camera in the scene has its own camera coordinate system, which is oriented according to the camera orientation).

Presence of positive Z coordinates of points in bundler.out doesn't violate the statement in the Bundler documentation you are referring to, as the points are stored in the world coordinate system. After transforming point coordinates to the camera coordinate system (as described in the Bundler User's Manual), Z coordinate should become negative (for points that are visible by the camera).

At the same time, neither format makes any specification on the world coordinate system orientation. Orientation of the world coordinate system can not be determined from the photos alone, and requires additional information such as ground control measurements. Thus, initial orientation of the reconstructed model produced by PhotoScan is completely arbitrary (i.e. undefined), and can be later assigned using Ground Control dialog.

[ju523m]

Hi Dmitry,

thanks for your answer. You explained that the model in photoscan is in arbitrary coordinates, unless it is assigned real world coordinates with the ground control dialog.
That brings me to the following questions:
- How is the origin of this arbitrray system determined / defined  in photoscan ( ... with respect to the camera positions / centers, e.g. center of first camera that is used for bundle adjustment) ?
- How is the orientation of the axis of this arbitrary system defined with respect to the camera coordinate systems, e.g aligned with axis of first camera that is used for bundle adjustment?

Kind regards