Workflow question with a huge Photoset

[Nedo]

[Andrea Alessi]

Hey Nedo,

it depends on the photos actually!

I personally prefer to run them in small groups and merge them togheter (either via control points or near camera). I throw them all togheter only if i am 100% sure there is a lot of overlap (i.e. video feed or a very controlled shooting pattern) but you can try anyways, no harm done.

You may also try to re-orient cameras by right clicking them.

If you can share the pictures with us, i'll gladly have a look at them and give less-generic answers

Andrea

[Nedo]

Hi Andrea,
I have some photos who dont align, is there a way to give this non aligned photos manually control points?
The photos are grey and i only can try to align them.

Thanks for any help!
Nedo

[Roberto]

Hi Nedo,

you can try to re-orient photos with the workspace automatically by selecting workspace - add photos and by drag and dropping the discarded photos. The problem is that if they were not aligned during the orientation phase, this method will hardly work as it's still automatic and more useful when you have completely new photos that you want to add. As a rule of thumb, it is better to try to run the orientation process again with all the photos from scratch with more permissive parameters if you feel that some 'good' photos have not been oriented.

Alternatively, if you have a region that has been completely discarded you can also proceed as follows. Start a new project with the photos that have not been oriented. You can then merge two workspaces by specifying a few common control points:
https://www.3dflow.net/technology/documents/3df-zephyr-tutorials/tutorial-6-workspace-merging/

[Nedo]

[Roberto]

Yes, if the cameras are oriented in a wrong way, you can also add control points from the images and then run an adjustment with the control points from tools -] workspace -] optimize camera parameters. This can help to improve the camera orientations and converge to a good solution.

If you would like to share the dataset with us, I'll be glad to test it to understand why you are having problems. I can create a shared folder on our server. For the first phase, we usually suggest to run the 'default' preset first and raise it up to 'deep' (that requires much more time) in case of problems.

[Nedo]

Ok Roberto, i will test it again tonight when at home, if the problem still exist i will call you
for the image upload. I may have to increase the brightness on the Photos, some are maybe a little to dark.

[Nedo]

Hi Roberto, i uploaded a few pics here for you (700MB Rar): http://uploaded.net/file/o9x1mcle
i have no success to align them, i tried merging two zep files, but the
result was not aligned well, what do you think about the photos? And what could be the problem?

I also have another question, when some photos not align they are greyed out,
what can i do with this picture group? Can i select the non aligned and re-start
to compute in a new Workspace.

Greetings Nedo

[Scott Nebeker]

I had almost the exact problem when I first started with huge datasets. Because of my own trust issues with photogrammetry software I didn't trust Zephyr to handle it all at once. I was working a model with 1500+ images. After I created a sort-of reference model, I began adding 100 images at a time thinking this was the best way of doing it.

To shorten a story that only proves my ignorance and stupidity: don't do that. You'll only lengthen the process considerably! Trust the software. If your shots have A LOT of overlap, I found that reducing the "ANNNumberOfFeatures" in the Samantha phase from its default of 500 down to 200 will help a lot as far as processing time goes.

[Andrea Alessi]

Hey Nedo,

i downloaded the dataset (slowly =) ) i will soon also contact you privately so that i can give you access to our private shared server, so you can upload more data there directly and will speed up the whole download/check process from us!

Andrea

[Andrea Alessi]

Hey Nedo,

i had a quick run. Using Urban/Deep all 49 images out of 49 gets oriented correctly.

Which settings did you use?

[Nedo]

[Andrea Alessi]

No problem!

The photos are okay there isn't much overlap but there is quite enough texture. Point is, in some pictures there aren't many keypoints, so you should probably either take a few more shots that "connect" the two acquistions. That, or increase the parameters personally i like to run everything on default on first try and then if some pictures don't get oriented, i try deep.

If you want to try directly in advanced ( i didn't ) i think the most significant difference in your case is made by the higher number of keypoints.

[img src="http://i.imgur.com/zDmJD5R.png" alt="" /]

On a side note, to see the presets differences, you can simply switch between presets/advanced with the desired preset selected!

[Nedo]

Good morning,
Andrea i experimented with deep settings and it works, but sometimes some photos are still greyed out. Still dont know
the next workflow if that happens, so i stopped Urban photos and doing smaller scans, need more training with 3DFZ.
(For tons of Urban Photos i'm back to another Photogrammetry Software until i learn 3DFZ better.)
But the 3D results are so good, i'm very satisfied I compared the 3D results with other Photogrammetry software,
and the 3DFZ scans looking best.

Greetings Nedo