I was hoping for some advice. I am currently working on a project where I am using an automated capture Rig to capture some less than ideal objects for photogrammetry. A fairly common request I am getting is some quite “flat” objects. I have informed my boss that this can cause plenty of issues during the object alignment phase. Unfortunately it’s not feedback that is listened too.
One of the main issues when conducting alignment on this sort of object I keep running into, is when capturing the objects on two sides. What can often occur is that either I get two components as the object is so thin I struggle to get tie points along the edges (think coins). the other issue is that with thicker objects rather than placing the camera positions on the two objects it “stacks” the two layers.
I was wondering if anyone had any advice for either of these issues?
Regarding the stacking issue, I believe this is because RC is seeing enough tie points to realise that these positions should be part o the same component, but not enough to align them correctly. I have already tried a number of standard fixes primarily around adjusting the alignment sensitivity and tie point settings up and down. Sometimes I can have success with adding tie points, but not always, and obviously this is quite time consuming.
If there isn’t a fix, does anyone know if my above hypothesis is correct, and if so would you have any documentation or a “techy” description of this. Only I often find my boss is more receptive to my feedback when I can point to other people having the same issue, or some “official” documentation.
Apologies I can’t share any screenshots due to copyright issues.
Can you at least post drawings of your automated setup and describe it a bit more in depth? It is hard to understand the issue, if we cannot see what your setup is.
For instance, if you are using some form of markers, are you leaving the markers as they are and then simply flipping the side of the object? If so, that would explain the alignment stacking those on top of each other, as RC alighns the images according to the markers that haven’t moved. That ore something like that could be the issue here.
But then again, it could be something completely different. We would need more information.
So it’s definitely on the alignment side of things, I’m a fairly intermediate photogrammetrist so it’s not the usual culprits, in object capture
Blockquote For instance, if you are using some form of markers, are you leaving the markers as they are and then simply flipping the side of the object? If so, that would explain the alignment stacking those on top of each other, as RC alighns the images according to the markers that haven’t moved. That ore something like that could be the issue here.
Specifically to the above idea I am using April Tags when on one side of the object and removing them on the other. Turntable and background are properly hidden so it’s not detecting tie points in the background (and I have also tried applying masks to the inputted images just to be on the safe side and that isn’t working either)
Have you tried to mount the flat objects on it’s edge, so that you can scan both sides in one turntable revolution? Doing this as well as adding april tags onto the turntable itself, should give RC more than enough information to align the images properly. Of course it is still possible, that the object might be too thin and produce holes when reconstructing the mesh, but maybe worth a try.
This will obviously only work on objects of a certain stiffness, otherwise you won’t be able to mount them standing on their edge without folding over.
