Any way to 'help' photo aligner?

[TijuanaKez]

Hi,

I've constructed some automated turntable systems with high accuracy capture data logging for another purpose and came across this program by accident.
I've been playing with the demo and the photo align really struggles with certain low contrast subjects.

With my system, the camera is on tripod, and the program logs exactly which angle each photo was taken in csv file.
But the camera's y position in the point cloud varies greatly, indicating an inaccurate scan.

Is there a way to inform the point tracker that the camera's y position is locked?

And for future versions it would be extremely helpful to be able to import the angle data somehow aswell.

If we know camera's y position and distance to object/zoom is constant, and we also know the exact angle of each photo, I would imagine this to eliminate high percentage of possible errors in the cloud data.

[Alexey Pasumansky]

Hello TijuanaKez,

Are you using masks for the this project?

[TijuanaKez]

Shot on a near perfect green screen and loaded in png sequence with alpha. Only a handful of background points were detected.
I haven't tried to use alpha as mask yet but when I get a chance I'll give it a shot.

So no way to lock y position?

[Alexey Pasumansky]

Hello TijuanaKez,

I suggest to import masks from alpha-channel and restart the alignment stage, since you've got masks anyway.
Also the problem may be related to the lack of texture pattern on the object  being scanned. Usually it happens for plastic or metal surfaces.

As for the camera orientation and position data, in principle, it is possible to load camera positions, but they should be accompanied with the camera calibration data. But I don't think it would work for the projects where the positions and angles are measured "mechanically", as the possible error would be quite big.