This last year, my colleagues and I at the University of Minnesota have started experimenting a lot more with reflectance transformation imaging, aka RTI. What is RTI exactly? Like photogrammetry, it works using many digital photos of a single object. That’s about where the similarities end however. When capturing photos for RTI, the camera and object remain stable. The scene is lit from one source, and the position of this source is changed from photo to photo. Software, such as RTIBuilder, is then able to generate a map of the surface of the scene, which can be used to create enhanced visualizations or to digitally re-light an object from different directions. In archaeology, this is especially useful for objects with features that are best viewed with raking light, such fine engravings or modifications on bones. For more general information on RTI, your first stop should be this page from Cultural Heritage Imaging. They have free and open source software that can be used to generate and view RTI files, and a ton of other great resources.
After working with Nadine Huber, Christian Hoyer, and Harald Floss from Tübingen last year on this paper comparing photogrammetry and RTI, I decided I wanted to make my own RTI rig. Many RTI rigs work by using a series of wired LEDs connected to a camera, which are programmed to fire in sync. These ‘dome style’ rigs are super efficient, and ensures even spacing between lighting positions. However, having not put together a circuit since middle school and with a budget of essentially zero to start off with, I decided to go for a more low tech approach. So one weekend last September I sacrificed a mixing bowl from the kitchen, drilled a bunch of holes in it, and bought a few $1 LED bike lights. This was the result:
Not too shabby! However, this rig had a few issues. For one, the cheap bike flashlights weren’t doing the best job of lighting the subjects. Also, the pre-drilled holes didn’t offer much flexibility. It was time to kick things up a notch. I took my inspiration from the rig created by my colleagues at Tübingen, which consists of a flashlight resting on a foam core arm attached to a wooden base. A foam core circle acts as a guide for moving the arm around the object at systematic intervals.
Last year, as part of a 3D printing workshop I toured the XYZ lab in the Art Department at the U of M. I remembered they had laser cutters! I figured I could design an arm that could be cut out of plywood or another rigid material, which would be durable, could be reproduced easily, and could be packed flat for travel. And so the silly prototyping began.
After tinkering a lot in Adobe Illustrator and a couple of trips to the XYZ lab I ended up with this:
I felt like we were starting to get somewhere! This is the version of the rig that we used to capture the images we used in our 2016 CAA presentation. However, I couldn’t possibly stop there! Over the next few months, I made a bunch of other additions / changes / improvements. In no particular order these included:
- Switching from 1/8″ thick material to 1/4″ material for stability. This meant I had to widen the cut outs where the pieces fit together.
- Making the arm shorter, since it’s not necessary to have lighting positions from so high up.
- Attaching the arm to a lazy susan bearing. This allows you to push the arm around the object being photographed instead of having to lift it each time. This led me to design a 3D printed mount for the arm, as well as a stable platform for the objects being imaged consisting of a 3D printed base and a laser cut top complete with angle markings.
Here it is in its final form:
Some of my results are below:
For archaeology, the applications for RTI seem almost endless! As a stone tool person, I’m especially excited by its potential as a tool for lithic illustration. If you’re into that kind of thing, you can check out many more examples of lithics images using RTI atcheck out my colleague Leszek’s site: rtimage.us).
Note that because the light positions are imprecise, with this rig the software needs at least one reflective sphere to be in each scene. RTIBuilder uses the position of a highlight on the sphere to calculate the angle of the light in each image. For larger objects, a lot of people use marbles or other spheres / balls covered in black nail polish. For smaller more zoomed in images, I’ve been using tiny silicon nitride ceramic ball bearings.
I also figured out that by combining a relatively cheap bridge camera with with high optical zoom with a snap-on macro lens (plus adapter), it’s possible to get really clear photos up to about 150X. Overall, the setup is cheaper than even a lower end Dino-Lite USB microscope, but is capable of producing much nicer photos. Combined with RTI you can get some pretty interesting results. Below is a video of a piece of bone used for flintknapping as seen with different filters in RTIViewer. The sphere in this case is 1/16 of an inch (~1.6 mm) in diameter.
I presented this rig as a poster last month at the annual meeting of the European Society for the Study of Human Evolution. I’ve put the poster, along with a full parts list / instruction manual for the rig, plus files for 3D printing and laser cutting on the Data Repository for the University of Minnesota. If you’re feeling more advanced and are interested in building an automatic ‘dome-style’ rig for yourself, I’d suggest checking out this set of award-winning Hackaday instructions from my colleague Leszek Pawlowicz. Hopefully, some of you will find these projects useful and be inspired to try out RTI for yourselves!