Sticker Problem

Got a bit of a puzzle. While reconstructing a set of buildings, I realized that my components are messed up. Cams are adjacent to ones they should be far apart from. After some digging, there appears to be some “rogue” alignment.

Anyone experience this before? Any suggestions on dealing with it?

Here are two images from two sides of the buildings (obviously not the same side):

However, RC thinks they are “matched” via an annoying little sticker on the doors:

Zoomed view:

What’s more interesting, there are plenty of other tie points in the images. But those are not used to prevent the images from aligning:

 

Any thoughts or advice are greatly appreciated. 

I had the exact same problem before, and for me the solution was to remove the close-up pictures of the identical object and re-align. I guess that there is some statistical algorithm that prevent this kind of false alignments for general case, but if you provide a closeup, then this is like if you provide a bridge to connect all the stickers together… 

I didn’t even do close ups. The close-up view is just a zoom of the original pictures.

I really don’t want to have to go and remove stickers from all the pictures :slight_smile:

Bit late, but that’s a typical false positive. Not surprising since the signs are identical!  :slight_smile:

What I would do is to align parts of the model where there is only one instance of those signs and then combine everything via components and “align components only”.

Hey Gotz,

Yeah I went the comp route as well. Sometimes don’t even have to restrict to “align comps only”. It just figures itself out …

Unfortunately, there are a lot of stickers. And RC also really likes fire extinguisher boxes, apparently :slight_smile:

No wonder, they are red and shiny! Not to mention really practical in the case of a fire…   :wink:

Anyway, I can imagine it’s a pain.

If you have access, next time you can try and put a few stickers (magnetic?) around those repetitive objects to sufficiently individualize them. Or you start using HUGE GCPs with an individual coding. Because to be honest if such a comparatively small area can align the whole rest of the image that sais much about the suitability of the remaining features.

Oh, and something else that could help is to start taking images from further away so that the containers and stuff are all on one image and small enough so RC can create nice features from them. If you then move closer slowly, RC should be able to sort them properly…

You should try out historic buildings once with nice irregular masonry…  :slight_smile:

I would love to scan rocks and old buildings and such!

Instead of flat, featureless, shiny metal panels