600 Boulders?

Doug Gann, Preservation Archaeologist and Digital Media Specialist

(February 28, 2017)—”Don’t bite off more than you can chew,” the old saying goes. In this case, to paraphrase another idiom, my eyes were definitely bigger than my computer’s hard drive.

The image above is a fully interactive 3D model. By clicking on the image, you can spin, zoom, pan, and just look around this ancient place. This model was created with the use of our “Gilacopter” UAV, which Preservation Archaeologist Aaron Wright used to collect some 1,025 images of the Painted Rocks. The 1,025 images were then imported into Agisoft’s fantastic Photoscan modeler, which used desktop or “soft” photogrammetry to build a digital point cloud of 256 million data points. This point cloud was then converted into a mesh model of roughly 100,000,000 polygon faces. From there, the model was archived at our data center, and then reduced to 300,000 faces so the model would be small enough to view on a webpage like this one. As you browse the models above and below, remember that you are only looking at only .003% of the available data.

With the primary model in constructed, Dr Wright and his team of volunteers began a comprehensive program to conduct condition assessments and 3D photoscans of each and every one of the roughly 600 boulders that have been modified by cultural activity both ancient (Patayan rock carvings) and modern (vandalism). The first set of boulders to be documented to date are the rocks that have seen the most vandalism.

One of our next questions was how we might balance speed and accuracy in recording these glyph-covered boulders. We had our standard photogrammetry camera, a Canon D30, which is an ideal camera for the photogrammetric process, because it can be reprogrammed by the CHDK camera hack, and records GPS data for each photograph taken. But we also had our new GoPro 4 modified with the Peau Productions 82-degree fixed focus lens. (Photoscan can handle the GoPro’s Fisheye lens, but anyone who’s been working in photogrammetry for any length of time is trained to consider fish-eye lenses as heresy.) Our problem with the D30 was that it turns out to be so very slow, only able to capture an image every 5 seconds, whereas the GoPro can record an image per second. So, we picked a boulder at random and our photography team tried out both cameras, which we again processed through Photoscan.

Canon Camera Test

The above model of Boulder 174 was created with the Canon D30, and comparing the output to the GoPro Silver 4 Camera below, most people see a more detailed digital model with the Canon photography, but a more complete model with the GoPro, presumably as the GoPro has the wider field of view. The “holes” in the models indicate places that lacked sufficient photographs to reconstruct a 3D surface. These holes crop up when it is not physically possible to squeeze a camera in-between an adjacent boulder, or a nice sharp set of cactus spines. Our field crews continue to utilize both cameras in the field to speed up documentation.

Beyond our hard-working, dedicated volunteers Kirk Astroth, Carl Evertsbusch, Fran Maiuri, and Jaye Smith, these three-dimensional web models are made possible by a service called “Sketchfab.” Sketchfab.com is like a giant 3D model bank that allows us to store models for public access. All the boulders we plan to model will be posted to a Sketchfab Collection, which will serve as a public repository for this digital data. You may access this collection at https://skfb.ly/66suU. So far, we have 20 out of 600 boulders modeled, and I’m going to have a very busy spring trying to keep up with our fantastic field research team.