Control Points going into wrong positions

Hi, we are trying to scale our model so we’ve started to add some more CP’s to the images however the controls aren’t aligning into the positions we have placed them in the images. Is there a reason why this is happening?

im2.JPG
< Correct one

im1.JPG
< Incorrect placed

Kind Regards.

Hi Dariusz Klimek

How much CP projection you have there ? that mean on how much images is placed one ControlPoint ?

Those controls are on at least 4 different images.

I have a similar problem. At first I thought it was just because my photos are bad, but look at this scene. How can point 21 be so far outside the passage?

EDIT: oops, messed up the screenshot, new one attached.

EDIT: new screenshot again. Was showing two green panes by accident.

Ok, I did an experiment that shows that in my case RealityCapture was not doing anything wrong.

Control point 21 had 4 images, but only two were from the current component. I took those two images and registered them together, and then the point showed up in the right place.

So I guess what was happening was that the other images that were used in aligning the cave wall “pulled” the wall inwards. The wall didn’t line up with the control points because the wall was calculated based on more images than the point.

Dariusz might have a different problem though.

Hi Dariusz Klimek

So CP its on 4 images and have at least 3 points visible on SINGLE image ?

Can post screen of one image so can see the overall placement of the CPs there?

Hi Aaron Curtis

Looking on your case the main issue is the CPs are too close together, you need to place them a bit apart form each other

Wishgranter… that might be the answer to some other problem, like “why are my components failing to merge” but I don’t see how it has anything to do with the issue in this thread.

I’ve had this a few times but narrowed it down to:
[list=][*]Too few images with CP in current chunk
[*]Misaligned image CP
[*]Distortion parameters not optimal (CP near edge of image)[/list]

The latter was mostly with fisheye images where the very outer edges of the image should be avoided for CPs but may be possible with “normal” lenses. Checking the tie points on the images with the CP may provide some clues if there are alignment issues throwing off CP locations