This week, I brought the Artec Spider 3Dscanner to the BVMC to try comparison with using an actual 3d scanner compared to using photogrammetry.

The week started with a trip to CHDR to get familiar with the scanner and its processes, as well as fill out some paperwork to be authorized to bring it off campus.

To run the Spider, it has to be tethered to a computer running a specific Artec software. So, borrowing the Spider also brings with it a borrowed (ancient) Alienware laptop with Artec Studio 17, the software it uses.

Broadly speaking, the workflow involves scanning the object, which takes some adjustment and getting used to. The scanner has a relatively limited field of view which is displayed on the computer’s screen as you work through the scan. It is sort of like you’re “painting” the object with the flashing light from the scanner – you need to get even coverage and not move too quickly.

For this, I again worked with the same turntable, allowing a slow, yet consistent, rotation of the object. For the first object, the canteen, we sort of “painted” up and down in vertical strips as the object rotated – if the movement were graphed against the object, it would be like a sine wave with a short period.

The first object we did at BVMC was this Canteen you may remember from a previous week. We struggled in photogrammetry – the object was a bit too shiny and the process never really got the details right.

Here, though, it succeeded in getting the details very well. After I was certain I had completely covered the object, we rotated it to get coverage of its base and then proceeded to do another series of scans. This was a mistake in our process, as it allowed the chain holding the cap to the body of the canteen to move.

3D scanning in this manner is different from photogrammetry in a few ways, but similar in another. One key difference is photogrammetry really doesn’t like it when you move the object – the process is designed to work backwards in reconstructing its model assuming that the object itself is unchanging. It’s one of the reasons that turntable based workflows, while effective, require some additional steps compared to methods that involve the camera moving instead. But it is a disaster if the object moves – the computer programs really get unhappy and it produces truly garbage results – because the data they’re working from is different. With this 3d scanner, because it is sort of working in real time and other technological differences, you can move the object while you’re working on it- in fact that helps it track when you’ve changed to a new side (as opposed to flipping the object while scanning is paused, it seems advantageous to flip it while the scan is ongoing).

however, we erred in re-scanning more of the object not just the base that was previously hidden. For many objects, this wouldn’t be a problem, but for the canteen with the very movable chain, this presented a challenge – as the chain moved!

You can see this in the model as there are sort of two different versions of the chain visible. I cleaned the model up a bit in Blender and Meshlab – if I actually knew what I was doing in either of those programs, I could go in and extract out the “bad” or extra bit of chain and leave just the portions I do want. But I don’t know how to do that yet – it’ll take many more youtube guides on how to use those programs before I’m able to do that effectively.

For model number two, we modeled a hat, purportedly from the border guards in the Soviet Union, though there was not much documentation to accompany it.

This model was quite tricky to make, though significantly easier than I believe it would have been via photogrammetry. It was a challenge to flip the object and maintain its presence in the field of view of the scanner. The hat is somewhat fuzzy and featureless in places (it seems to be felt or felt-like). When flipping the object, it was helpful to keep the more detailed portions of the object in the scanner’s field of view – like the buttons on the side or the crest on the front.

It was challenging to “make the turn” and go from the outside of the hat to the inside of the helmet, because it is so thin it was relatively challenging to ensure that the scanner could “see” that area and begin to make the transition down into the interior – indeed there are numerous holes and other issues with the inside of this model. I’m pleased with it for a first attempt – but it would benefit from cleaning up in Blender or Meshlab. The inside is relatively featureless, so it would be well suited for reshaping digitally and blending the textures. There is also a sort of round/hemispherical protrusion that is on the top of the hat in this model that does not exist in the real thing – I’m not sure exactly what it is/where it came from, but it materialized at some point along the way in processing and I was unable to edit it out.

Interestingly, it did not materialize when I re-ran the processing in a demo version of Artec Studio 19 on my powerful home desktop, though I cannot export anything out of the demo version. I think I can borrow a software license from CHDR at some point and have the full version on my computer. A project for another time.

We had difficulty when attempting to scan some smaller things, like medals (a purple heart and a medal of honor) that seemed too small/flat for the scanner to pick up, or at least to be able to scan and then get the reverse side. I’ll watch some more videos and read more white papers about processes. Next week, I’ll head back with the scanner and attempt some more objects.