Special features for turntable setup

[mr_cg]

I am doing some experiments to capture 3d mesh by the usage of a turntable.
Depending on the object it works pretty well. :-)

Anyhow ... after some tests i would like to ask for some special features for the usage of a turntable to make it more comfortable.

Turntable Features:
1. Align fotos by pairing the cameras based on the filename sequence order AND pair the first and the last file in the sequence
2. Predefine the rotation degree per camera. This should speed up the alignment and improve the accuracy. It should be also possible to handle row sequences taken from different angle/high.
3. Show the aligned cameras (optional) in the workspace
4. Import Mask from "Green Screen" or "Blue Screen". I am using "from Background" which works pretty well. Anyhow greenscreen/bluescreen support would be also a great option.

[mr_cg]

Sorry, forgot to mention the most important feature.

As already mentioned i am using the import mask "from background".
1st i make on picture from the turntable witout any object. All background is green.
Then i start the sequence with the object on the turntable.

This works really great but i always need to select the "background" picture for every single picture.
Unfortunately it is not possible to select one background image for all cameras in the workspace.

Because of this i would like to request multiselection support for the "from background" feature.
One background image for more/all cameras.

More me this would be the most important missing feature.

THANKS

[Alexey Pasumansky]

Hello mr_cg,

Thank you for the suggestions.

1. It is not the most effective way, since oblique images may have overlap even it they are opposite on the "shooting circle". Another problem is that such approach prohibits image alignment from several heights in a single run.
However, in Professional edition such pre-selection scenario can be simulated using fake ground control data, for example x and y coordinates are calculated as a position of the point on the circle with radius = 1.

2. I'm not quite sure what do you mean by this point. Camera orientation angles (yaw, pitch, roll) are not currently utilized in processing.

4. You can create an image file with the same dimension as photos in the chunk are and fill it with any color you like, then just use this file as a background import file. In Professional addition it is possible to specify the color of the background from Python script when mask from background feature is used.
And as for your last comment, I think you should use templates for specifying background files, for example, using {filename}_background.jpg in the "Mask file names" field will allow to use IMG_0001_background.jpg file for IMG_0001.jpg camera and so on.
However, omitting templates you can just specify background.jpg static filename and this file will be used for every photo in chunk/workspace.
Is that the requested feature?

[mr_cg]

However, omitting templates you can just specify background.jpg static filename and this file will be used for every photo in chunk/workspace.
Is that the requested feature?

This was exactly the feature i wanted. Thanks!!!

Coming back to the request 2.
The reason why i am asking this is because sometimes the automatic alignment of the cameras are completely wrong. E.g. picture is aligned 180° to the oposite because front and back of the object on the turntable are quite similar. This could be prevented if i could tell the SW that e.g. pic1 is 0°, pic2 is 15°, pic3 is 30°, pic4 is 45° etc.
Based on this information the SW can ....
a) limit the automatic alignment only to useful combinations (compare pic_30° with pic_15° and pic_45°, but NOT with pic_210°)
b) use this information as additional information for valid/unvalid points. If the SW realizes that the rotation should be e.g. 60° between two cameras while user has specified the cameras should have difference of 15° rotation ... then photoscan can imagin that the alignement does not match.
c) i am not sure about the algorythm but it could be possible that finding matches could be speed up in case the rotation is known. The SW do not need to calculate possible matches over wide area.

This option could be similar to the (free) hugin stitching sw.

Many thanks for your support.
Photoscan is a really great SW.