I’m interested in having my camera circle an object in 360 degrees. Is there any technology out there designed for this situation? I was thinking of making something myself, but the designs seem rather complicated. The goal is to reconstruct a 3D model of the object. I’d like to create a large dataset of such models.
I’m not sure what your final goal is, but it might be easier to put your object on a turntable and rotate it 360 degrees.
What are you trying to accomplish?
indeed easier to keep the camera stationary and rotate the object if you can.
Turntable
White fabric curtain hanging from a Hula-Hoop, suspended around it, with a slit on one side for the camera lens to stick through.
Lights above, and on the fabric from outside.
Turn, click, turn click, etc.
Tris
Unless (the reason I asked) the OP is, for example, doing this in such a way that he needs that background in the picture as well.
Also, if you do need the camera to move, how big are the objects that it needs to go around and how smoothly does this need to happen? Does it need to happen with electronics and motors for a very smooth action? Is it okay if it’s something that you simply adjust for each picture? Is it okay if it’s something that we think up that’s being repurposed for this job. That even though it’ll still work, might not give optimal pro results?
ETA, is there a budget to go along with this project? I assume setting up a track and dolly would be too expensive?
I can imagine spending 2-300 dollars on parts. The software is pretty robust to irregularities so it doesn’t need to be very precise, but ideally I could move the cameras at small intervals if desired.
I like this solution, except I think it needs one modification. The turntable should be able to move up and down as well. I think there should be at least three vertical camera positions, above the object pointing down. Level with the object pointing straight. Below the object angled up. Still pretty complicated…
If you can suspend the object from some fishing line to get the “below” shots" you should be set.
ETA, in fact, you may be able to use the fishing line for all the angles, if suspending the object is possible.
The OP didn’t ask, but don’t count too much on the 3D models you can generate this way. It’s still more an art than a science, and there’s only so much software can do. To get good results, you’ll either have to get really lucky, or adjust the final model yourself.
Actually the software is pretty good, particularly if you give it a clean, well constrained dataset. I’m looking at Bundler + CMVS + PMVS2.
Or,
make a elevated stand out of a wooden disk with a nail tripod. Center it on the turntable, cover with the same white cloth, and place the object on it. Now shoot the shots from level, and then a set from up, looking down, then turn the object over, and shoot again looking down, digitally rotate the second set of shots.
Tris
I think lasers rather than optics would be necessary to make a truly accurate 3D model. I watched a show on how they made 3D models of European cathedrals with lasers and could tell which parts of the supporting structure had moved and how much.
Shouldn’t make much difference - it’s only trianguation, and light rays from lasers aren’t any straighter than rays of ordinary light.
True, it’s really all optics, but lasers are more precise.
Lasers allow you to triangulate and capture points in real time, but if the same features/points can be identified in a pair of photos from different angles, it should be possible to perform exactly the same triangulation. Photographic capture could actually surpass the laser method, in some circumstances, I suspect (it also allows you to capture the texture map in colour at the same time - I’ve seen this done with a three-colour laser system, but the results were less than impressive.