I’ve spent a lot of time at the AMNH, considering the fact that I have resided in Europe all my life. Discounting a visit when I was ten years old, it’s been four visits totaling roughly five weeks, I guess, most of which was spent looking at dinosaurs. Both in the exhibitions and in the collection rooms.
Today, I happened to comb through some of the folders with pictures from these visits, when my new-found passion for digitizing via photogrammetry piqued my curiosity: how much ‘accidental’ 3D information was in the pictures I had taken in, e.g., 2003?
Obviously, I had taken loads of pictures once I had a digital camera available. They cost you practically nothing, so taking more is an excellent idea. However, I had obviously not tried to take photos suited fro photogrammetry, instead preferring the usual exactly later, exactly frontal, exactly dorsal and so on shots that one wants for creating figures for papers. However, I had often taken oblique shots, and the glass cages many specimens are in sometimes made orthogonal shots impractical in the first place. I did chase good old (not really) Carl Mehling around the AMNH dinosaur halls one day with a big piece of plywood, trying to block light sources. That helped some, but wasn’t neither perfect nor did I dare bother him too much. Also, Carl had two interns back then, and they laughed themselves silly at the occasion, especially when Carl was pulling silly faces, which did not help me with having a steady hand.
So I started throwing photos at my photogrammetry software tonight, and here are some of the results:
Golly! That’s half a skull from 14 photos!
Note that I didn’t fiddle with selecting suitable photos, masking them, or anything else. Just straight import, align, mesh. Now, after the obligatory theropod, some “proper” dinosaurs.
OK, this pelvis in the Big Bone Room didn’t come out quite as well, but the ilium is very nice, the height of the neural arches well documented, and the width across the tips of the ilia as well as the acetabulae is easy to find: measure from the midline (based on the neural spines), then double.
The next one clearly needs some work – some photos do not align with the others. That’s probably (most certainly) caused by the object being moved versus the background in between photos. I’ll have to try some masking of the background to fix this one day.
And….. more theropod – this one surprised me a lot – many of the photos were taken in 2003 with a pretty bad camera.
the weird boxes you see are only in the 2011 photos: digitizing equipment. More on that later.
But this doesn’t always work this well. Check out what happens when the object of interest was moved between shots! I didn’t realize this from the thumbnails, because the background was fairly light. I’ll have to try this again with masks, so that alignment is dependent only on the pelvis, not on anything that happens to be around it.