Agisoft PhotoScan

Agisoft PhotoScan => General => Topic started by: jak on November 20, 2017, 05:46:17 PM

Title: spherical model
Post by: jak on November 20, 2017, 05:46:17 PM
Hi,
I have been experimenting with the building models from spherical video. So far I have had great results.
However, 1 scan has been giving me problems.
The scan (a long scan of a back alley about 150m using 800 photos) works well all photos are aligned and resolution is great but the model/alley get smaller and smaller from 1 end to the other. As if the space shrunk as we mapped it :)
In the two screen shots one can see how the alley gets smaller and smaller to the left.
Any ideas how I can resolve this?
Title: Re: spherical model
Post by: jonny5alive on December 08, 2017, 02:29:03 AM
Did you get anywhere with this?

I never got anywhere with spherical which is a shame. I'd love to make 3D maps from walking through a scene.
Title: Re: spherical model
Post by: jak on December 08, 2017, 02:48:46 AM
Some really mixed results. The model in the photo is really really good, apart from the problem stated!
And the capture work flow is great, one slow movement through a space.
will post some more pics soon.
Title: Re: spherical model
Post by: jonny5alive on December 08, 2017, 03:01:42 AM
That`s good to hear.

Have you thought of putting some markers in that will help PhotoScan calibrate itself, perhaps on points that have known measurements?

This was interesting to use and you might want to try it out:
http://ccwu.me/vsfm/
Title: Re: spherical model
Post by: Alexey Pasumansky on December 08, 2017, 09:14:45 AM
You can check the following articles related to the processing of spherical panoramas (from video sequence) in PhotoScan:
http://journals.bg.agh.edu.pl/GEOMATICS/2015.9.2/geom.2015.9.2.51.pdf
https://www.isprs-ann-photogramm-remote-sens-spatial-inf-sci.net/II-5/211/2014/isprsannals-II-5-211-2014.pdf

Optimizing the alignment using GCPs, know camera locations or relative distances should help to scale the model to real dimensions.
Title: Re: spherical model
Post by: jak on December 12, 2017, 12:03:48 AM
Thanks Alexey,
I had a quick look over those texts already. Will study in greater detail.
I'm right in thinking that using GCP's would mean re shooting the footage?
Or can I use markers in post?
Thanks,
J