Should poor areas be trimmed off each chunk before merge chunks ?

[Steve003]

Hi,
we all know those areas that look like the base board has a snow drift up to the model, covered in texture thats gone blurred.

I have a metal object photographed on a table. laid on its side then standing upright, so  have two chunks 1. top and sides, 2. underside and sides,

can I just go ahead and merge chunks, after first bringing them into the relationship they are, which in effect will see them intertwined with each other, and expect merge chunks to remove the not so good bits, or do I have to trim off the parts the other chunk will provide and repeat such trimming on the second chunk ?

What is the correct way ?

Steve003

[Landy]

Why the two chunks?  Have you tried aligning all your photos as one chunk?  I find it usually works without having to split the top and bottom photos into separate chunks, depending on what the object is.  When I have to split into chunks, yes, I do trim the white areas around the object, but I have no idea whether it is strictly 'necessary'.  Probably depends on how well the merging goes.
cheers
Bill

[Steve003]

Hi,
Such wouldnt work.
The item was sat on a picnic table covered in flat newspaper which was taped down , each time the item was moved to photograph its underside or topside etc (called a set for each photo set) the newspaper was then different to the item.
I happened to make a mistake and fed two sets to metashape at the same time, and when align photos was done I had a double image going on.
So I had in fact done what you suggest by accident.

If I were to spend many hours masking out the background and newspaper from a few hundred photos then yes that would work. I dont have time to do such, as this is one of many projects, and I am years behind as it is. That metashape is often failing to align a GOOD set of images and photoscan manages it perfectly everytime is also a real pain as it messes me about. Wish Agisoft would solve that failure with metashape, they are aware and have photosets that prove it.

As such the newspaper helps as it provides a nice detailed part of the subject to ensure good alignment, bit like targets but far better.

That it keeps reappearing despite trimming it off in initial cloud after align, as I am using mesh from depth maps, IS DOING MY HEAD IN.

another issue Agisoft should really fix as its a fundamental flaw.

Anyhow it sounds as if I should trim away the model (I have each set as textured now) leaving only quality areas to merge.

Should I take the shaded stage and trim and merge or take the final textures stage and trim and merge ?

I wondered if 'merge chunks' was 'intelligent ' enough to merge the best parts and discard the not so good parts, lessening our workload.

Steve003

[PolarNick]

You can try to mask a newspaper (to align all photos in single chunk) with just a couple of masks like shown here - https://agisoft.freshdesk.com/support/solutions/articles/31000158967-aligning-photos-with-background-suppression-from-single-mask

Or in two passes - generate target model in each of two chunks - delete background triangles (a newspaper) - generate masks from model. After that you will have good masks for all images that will make possible processing of all photos in a single chunk - https://agisoft.freshdesk.com/support/solutions/articles/31000163388-automatic-masking-from-the-model

[Steve003]

Thanks, I will follow those through.

Steve003