Photogrammetry is perfect – if everything work fine

That photogrammetry, the method of calculating 3D representations from a series of 2D photos, can be a great tool in palaeontology is quite obvious when you take a look at Peter Falkingham’s paper in PE (and the many other papers on the method out there). But just how good is it? I’ve previously shown results from a free service on the net, and from my own photo tour of the Stan T. rex skeleton in Brussels. Today, I want to show you sometimes, things can go so wonderfully right that there simply is no better way to digitize a bone than photogrammetry.

Here’s a photo of a scapula-coracoid cast of Giraffatitan on exhibit in the MfN Berlin.

As you can see, I was able to have it pulled into the center of a large space, so that I was able to take 111 photos (yes, that many!). I made sure to get shots in circles, moving horizontally about 10° between them, one circle from just above the ground, one from shoulder height, and one standing on top of a table that I dragged around the bone. Additionally, I took some shots from closer in, trying to capture the down-facing parts of the glenoid and so on. A handful of photos were slightly out of focus, so I deleted them.

And this is what agisoft Photoscan and about 30 minutes of editing spit out (vertex color removed, to make the quality of the surface better visible):

Oh, I should note that this is 10% of the original file size, too :) There were a few holes to close, but very few issues with sharp changes in geometry. Thus, the editing was very easy. I wish all my attempts had worked out so well. Check out what became of my attempt to use photos of the mounted Giraffatitan to get a 3D model of the humerus:

OUCH!
And yes, the lower side is fine, but this side is concave instead of convex (under all that rubble).

About these ads

About Heinrich Mallison

I'm a dinosaur biomech guy working at the Museum für Naturkunde Berlin.
This entry was posted in 3D modeling, Dinosauria, Giraffatitan, MfN Berlin, photogrammetry, Sauropoda, Sauropodomorpha. Bookmark the permalink.

6 Responses to Photogrammetry is perfect – if everything work fine

  1. Stu Pond says:

    That is an excellent mesh. I haven’t tried doing a complete 360 yet, now might be the time.

    • To be honest, this one did not work too well the first time around, with the exact same photos and settings (which leaves me quite confused): the first attempt was unable to distinguish between front and back, so that a single surface of zero thickness was created for most of the blade.

      I’ll post some more stuff like this soon, highlighting some of the problems I stumbled upon – and solutions for them, if I found any.

      • Matteo says:

        Tried with different chunks and later a new or manual alignment? it worked with a sauropod vertebrae made with less than 40 photos and without tripod.
        I like “easy” things

  2. SanSaurio says:

    I like a lot this images. Congratulations!!!
    Whit this method, do you joint two diferent images in one (for example two vertebres scanned in two independent files and then joint in a file)?

    • Well, once this gives you a 3D file, you can just take it into a CAD program and do whatever you like. But you can even join files in this program, if there is overlapping areas (background is sufficient), by combining two chunks after 3D geometry creation.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s