Spherical images in PhotoScan Pro

[pbourke]

For the last couple of days I've been processing various image sets from a 360 camera roaming around a house. The testing is mostly around trying appropriate frame rate samplings from the 30fps video, every 10th frame, every 5th frame etc.

What seems consistent is that I get a reasonable and consistent number of tie points across all the images in the sequence, BUT only a subset of the images are "green ticked" and go on to contribute to the 3D model.

 The sets that do contribute are all consecutive  ... but then just stop, even though the images and number of tie points don't look any different.

Ideas?

[Circuit]

Ideas?

My guess is that you have a very large amount of PhotoScan (360deg) data/resolution in the Horizontal plane.
And a very small amount of PhotoScan (360deg) data/resolution in the Vertical plane.
i.e. At any series of camera  point-of views, and therefore PhotoScan point-of views, the horizontal
and vertical resolutions are very different, because the camera is only moving Horizontaly.

Again my guess is this is a common problem for non-360deg PhotoScan users attempting to use only
a one camera rig. i.e. Once Again at any given camera point-of-view, and therefore PhotoScan's 
point-of-view, resolutions the horizontal and vertical resolutions are very different.
Typically, with this case, the camera or subject is then only rotated in the horizontal plane.

My Question's, for both 360deg non-360deg multi-camera rigs, are:-

a/ Is there a optimum and/or minimum ratio for Horizontal verses Vertical  PhotoScan  resolution.
b/ Would various triple-ganged multi-camera rigs have advantages over a vertical dual-camera
(i.e. a stereo-pair) rig.
c/ Or would the use of dual-camera rigs and  triple-camera rigs have no practical advantages.

@ pbourke are you located in Western Australia ?.

[Kiesel]

Maybe you are interested in this paper: https://www.int-arch-photogramm-remote-sens-spatial-inf-sci.net/XLII-2-W3/85/2017/isprs-archives-XLII-2-W3-85-2017.pdf

Regards