im doing my first experiments with a church and have some questions.
18 mm, f/11, Automatic shutter and tripod
as you can see I went alongside the walls taking a picture towards the other side every 25cm. (maybe to little distance from one shot to the other?)
I’m pretty happy with the alignment - didn’t lose any picture here.(220) pictures,
however, the model is 44mio tris large and has 141 parts.
if i look at the walls, the geometry is quite bumby.
I can simplify the model - however it kind of affects everything and if i go down too much it hurts the model noticibly - bumps in the wall still there though.
So my Question is: is there a smart way to improve the geometry - maybe even simplify it without hurting the overall appearance ? The walls could be represented much simpler i guess - but how to arrange that ?
that is a very good result for the first try, with all cameras in one component - even for the second or third…
Anyway, I guess that the walls are pretty unicolor and not much texture, right? In that case there is not much you can do apart from going closer until there is some recognizable texture there.
I’m not so sure that the distance is the cause for the noise, but you could find that out easily by using only every second image and start a fresh project with those. What also improves geometry a lot is if you do another round at a different hight, so you get more stability. It might also help if you turn the camera 90° for that round, especially if you use exif grouping that can help quite a bit (from what I’ve read, that’s standard for paintings).
Hi Götz
Thanks for the advise. I actually tried with half of the photos (50cm steps instead of 25cm) - much wors.
And I totally agree for an anfternoon Test with 200 photos I’m pretty happy. - now I’m just trying to find out wich Parameter to tweak to get even better
I’ll try both of our suggestions : an additional round with different angle and detail shots of the wall
Thx
Regarding detail shots of the walls: Don’t just have the shots from all the way back and the detail shots but also take some shots inbetween to give RC the chance to connect those photo sets. Kind of “zoom in” in steps. Preferably from multiple positions.
Otherwise you will get multiple components and will have a hard time getting them to align all together.
For a situation like this, If I know the walls are flat, i would model all the flat surfaces in a 3d modeling program and merge that with the other details from the scan. There are lots of posts related to this topic but it also requires a lot of knowledge in 3d modeling.
I find that works best, especially if the surfaces are shiny or have a very flat color making it hard for the software to reconstruct.
