Tornado effect or BS

Hi,

It took me several trials to get the calibration to work out. But despite successful calibration, I have not been able to run a single correct scan with the turn table. 'Diet' pointed a tornado effect on the KS website and a youtube video from the Eora guys. But I always get a tornado effect, even when I'am sure that I have moved nothing during the prescan and scan steps.
Here are some examples:





If anyone has a clue, I'll take it!

Best

Hello,
I can’t get rid of the tornado effect, or ghost dance, or Indian dance effect either. Sometimes it’s very bad, sometimes acceptable. If you know any software that can stitch everything back quickly, I’d be glad to try it. With Meshlab, it is still quite tedious. But if scanning, stitching, and cleaning meshes takes that long, then it’s problably best to learn how to digitally resculpt things from scratch!

My feeling is that it is related to a wrong estimate of the distance between the phone and the turntable since the Indian dance effect (which is a better name than the tornado effect indeed) can be attenuated or worsened when the turntable is placed closer or hold off the scanner (in which cases the initial green light is not anymore centered on the central cross of the turntable).
This being said, might be worth to try without my iPhone protection case since this might slightly change the distance between the scan and the turntable.. Can the users that have got successful scans with the turntable tell us whether they removed their iPhone protection case?

I always remove my protection case, but I can’t get super clean scans. I noticed that my turntable doesn’t come back exactly at its initial position after a series of scans, it goes 359 degrees or so, so that the blue line is not aligned any longer with the white markings.
Mystère!

The exact distance between scanner and turntable seems to be very important to get perfect scans. So we could try to find out what distance is optimal.

As Totoleheros and Diet99 suggest, I suspect the problem is the scanner not correctly knowing the distance to the center of the turntable.  (Presumably, the folks at Eora 3D designed the turntable so its rotation is repeatable, consistent and known, and communicated properly to the scanner software.)

How about placing a simple, pointed, reference object (e.g., a cone or a rectangular pyramid) such that its center point is directly above the center of the turntable.  Perform a series of rotational scans with the turntable at different distances from the scanner along the alignment ray.  When the actual distance between the scanner and turntable matches the value used by the calculations, the scans should converge, and the point at the top of the object should align into a single point.

Once you've located the distance, as long as you don't bump/move the scanner and turntable, you should be close for scanning your real object.

You should be able to use the low-res scan mode and a small number of rotation slices for each step of the turntable placement experiment.

I would try this myself but have neither received my scanner (DHL says they'll deliver it Monday, 1/22), nor do I have a compatible device.  Next weekend, maybe...

aae83

One more thing to try for aligning scans: make a fixture that always provides three distinct locations in space from which a reference plane and location can be derived.  Make that fixture part of every one of your scans (e.g., set your subject atop the fixture), and you should always be able to manually align the results, using one translation and two rotations.

My thought is to make a platform having at least three point references in view at all times - for example, a hexagonal platter with raised points at the corners.  If three consistent points are always in view in your scan, you can use your 3D software to align two of the points (the first by a translation; the second by a rotation), and then rotate one of the point clouds about the line defined by those two points such that the third point is brought into alignment. The two clouds should then be aligned, as well as they're likely to be.

It's not as "simple" as having the device and software do all the work for you but, clearly, the software is not capable at the moment.  This solution, if not exactly simple, is at least straightforward.

This solution should work whether or not the turntable is centered at the best location, provided at least three distinct points in the two scans can be aligned to identify a common reference plane.

Cheers,
aae83

It is a good idea indeed to set reference points. However I was expecting a simpler use with less set up than in photogrammetry. That is the whole point of having a laser beam, just getting a single single line of concentrated light.
If the distance between the turntable, the camera, and the laser tower cannot be determined, then what’s the point of the calibration manipulation? Magic incantation? By aligning the blue line-white markings-cross-green laser line, and a given distance by the arm mount, don’t we have all the magic of Al-Kashi’s Law of Cosines and a known triangle?
I grant you that according to the phone model or if it is in a case, then we can’t really know the distance between the lens and the laser optic.
Somehow, I think they should have given an option to manually enter parameters, for example to define a bounding box of what is to be scanned. I have tried scanning with a background for better contrast on the object. Sometimes the background is captured too, sometimes it is ignored. When everything happens automatically, then it is difficult to know what to do to get better results.
I’ll try again with the rpm set on slow, and tinkering with various parameters.

I believe that main problem is that 3D trigonometry of an organic shape such as a figure or small statue requires calculations in a non-Euclidean system. You can probably scan a dildo ( excuse my French, I am French), but it’s not the kind of toy I have in mind to use.

I have 4 scans in a row of the same object.. all tornado effect. I did not move the phone on any of the scans. very frustrating.