Photoscan rocks my colletion digitizing project!

[Heinrich]

Just wanted to point out to all that I am using Photoscan Pro for a collection mass digitizing project, in which I expect to make high-res models of some 400 dinosaur bones within 5 months - and Photoscan absolutely rocks!   

The bones are between half a meter and 1.5 m long, and I take roughly 200 photographs per bone. Half with one side of the bone up, then the other half with the bone flipped upside down, and laid on a different table in a different corner of the room. This way, matching points can only exist on the bone itself. And I have a ratio of successful models of 100%! It ALWAYS works!

Then, build a dense cloud, clean it up for a minute or two (there is always a bunch of points floating around from the backgrounds), build the mesh - DONE! All in all I spend less than 15 minutes per bone, for models with an accuracy of ca. 0.1 mm and around 8 to 20 million polygons each. Soon, the models will be accessible online, a huge boost to our science.

Well done, Agisoft team, and thank you very much!

[eduardoestrada85]

How mesaure the accuracy?

[Heinrich]

- several scale bars, difference between them in Photoscan
- measurements done digitally and on the real fossil compared

[Kiesel]

Very impressive project, digitising 400 dinosaur bones with Agisoft! Waiting for the first bones online  .
How long does the processing take for one bone in your PC ? Would you say something about its hardware configuration?

Who wants to read more on this, can follow Heinrich on https://dinosaurpalaeo.wordpress.com/.

Regards Karsten

[SaussageDog]

Thanks for this info Heinrich. I am writing a dissertation on the use of photogrammetry in archaeology and will almost certainly be referencing your work and your paper in my discussions. Pleased I found your post !

[Heinrich]

How long does the processing take for one bone in your PC ? Would you say something about its hardware configuration?

It really depends on CPU, GPU, and number of images. Between 4 and 12 hours per bone, on some really nice machines with 8 to 12 cores.

[ThomasVD]

Are the tables filtered out automatically (since they're only recognised in half the images) or do you have to clean them out of the point cloud manually ?

[Heinrich]

Are the tables filtered out automatically (since they're only recognised in half the images) or do you have to clean them out of the point cloud manually ?

I need to clean them out, but usually that is quick work. There normally is a small gap between the bone and the supporting material. It is just a matter of rotating the dense cloud so that the view looks right at the gap, then lasso the unwanted points and delete them. 

[ThomasVD]

Ok cool thanks, might be good idea to put the bones on small supports from underneath then, and leave some space between the bone and the table. This is also really useful information for our discipline (maritime archaeology) where we often have to digitize a lot of ship timbers. So thanks for sharing!

[Heinrich]

might be good idea to put the bones on small supports from underneath then, and leave some space between the bone and the table.

that's pretty much what we try to do! We use Ethafoam which is cut as a tapering pedestal, so that the bone overhangs the support a bit.

Btw, interested in coming to the follow-up conference in 2017: http://www.naturkundemuseum-berlin.de/forschung/tagungen/digitalspecimen-2014/home/?Fsize=0&Lightversion=0

[ThomasVD]

sounds interesting! But 2017 is a long time away haha, keep us posted!