Getting good scans (Page 3 of 3)

Arnaud, great Vaas bobble-head mashup!

Check the dimensions of the scan results.  If it is wider than what it is supposed to be but not shorter, then it suggest that your camera is not centered on the turntable or that your lasers are not oriented to the center of the turntable.  If it is the proper width but is short, then it suggest that your camera is tilted up and not looking straight ahead.

- Uriah

Hi! Thanks a lot for your feedback. Yes, my Vaas-vahiné has been used as a model for my Atlas 3D tests :-).

After your remarks and suggestions, I checked the global calibration, and after few tries, everything come to a correct scale again. Now I'm encountering other issues. Globally, the scan results are nice, but often too messy, and I don't understand why. I tried different situations (different light expositions, different backgrounds, only one laser, default resolution instead high, ...) and at the end, I'm not able to find a common mode.
For example, I tried to scan a Lego Omer:

With both lasers and with default resolution, I've got the folllowing result:

With only one laser, I got this:

When trying the high res and both lasers on those two models:


I respectively have the following results: blurry, unfinished and asymmetric:

The worst result is on the S. Gainsbourg head. I never succeeded to scan it properly. Never. Is it due to the fact that it is a bit too small? For the other models, sometimes it is correct, sometimes not. I also tried to push some more light on the model to scan, and I think it is worst than only natural daylight.

Just for info, and to give some context on my scans, I took some inspiration from this forum concerning the calibration and built this (used on top of the small printed device for camera calibration):

Also, for the correct superimposition of the two lasers, I'm checking it like this:

So, if I'm not wrong with the calibration (and I think I'm not), I'm then lost with the scanner parameterization and environment (mainly regarding the model exposition).
Does someone have some clues for me?

Exact, the main source of light is coming from the right (due to the way the scanner is installed on my desk)... This could lead to part of my problems. Concerning the reflection, I've tried to expose my model(s) with direct front light exposure (halogen lamp). It seems it was not the best idea as the result was even worst. Should I use low power LEDs (as Pirvan does)?

The above comments are right on.  Also, it also looks as if the model may have been sliding around a bit.  Make sure you use the table liner to prevent the model from moving.

wow this is exciting seeing not only the support on this scanner here but whats possible with it and being as open source as SD printers etc for tweaks/mods. cant wait til I can buy one.

think I will make a tent to use it in for better control of light also

PS like homers halo lol

Thanks a lot for your help.
I did installed the latest software version, and ran the auto-calibration in addition to the calibration grid. I didn't used the table liner as I thought my small support (in fact, a foam cube) would be enough. I couldnt't figure out that the model was sliding on the table, I mean, that a microscopic shift could constitute a major issue for the scan... I'll definitely try the liner.

I'll do a lot of other tests, for sure. I'll buy also a RGB LED strip to be able to change the light, and I'll try to resorb reflections with talcum or tanning powder. I'll keep you informed. Once again, thanks a lot fo the help.

Pirvan, what are generally your parameters for your scans? High res and both lasers? What threshold for the laser? Only one table revolution?

Am I missing something? I thought scanners where designed to see only the laser light or whatever emission it is using to scan with. The cameras or sensors are supposed to be fitted with lenses that only pass the reflected laser light. That being said ambient light should have little to no effect on a scan. What I see in these scans are shiny spots on the models causing spurious reflections that the camera is picking up and causing the bad and or distorted scans.

If these scans are mostly for printing, why does it matter if you have color data? Most can only print in one color anyways. I would sacrifice color for precision any day.

If you put a band pass filter on your camera that passes only 365nm (Correction, should be 635nm) or 640nm depending on the type of laser diodes being used you would get no noise from ambient light reflections. That frequency is found in very few sources of natural light and is so narrow and of low amplitude it is nearly undetectable by cameras with band pass filters. Only lasers can generate a sufficient amount of it to be detected. I did holography for five years and am very familiar with that small line of the visual spectrum.

Hi all,
I tried some scans this week-end (taking into account all the abbove remarks), and the results were ... disparate. My calibration was perfect, I tried several parameters, and my conclusion is that the light exposure is the most important thing: too bright leads to a lot of artefacts, holes, blurry parts, ..., and too dark is not ok as the PiCam results are quite bad in this kind of conditions. Natural light is the best, but not always available :-). So, artificial light is necessary... and the problem is to find out a good one, and a good installation (should be indirect, and mat).
Other interesting observation: my scans with default parameters are seeminlgy better than in high-res (same model, same moment of the day and calibration verified as ok). I don't understand why.

Anyway, I never succeeded to have a high quality result as you can have, Pirvan. The last one is almost perfect, well done.

One more remark: calibration is difficult, and not permanent. I have to re-calibrate the lasers after one or more days without use. I suppose this is due to ambiant temperature variations. But it quickly becomes an easy thing to do :-).

I'll try to continue to give a feedback as soon as I have made some progress with my Atlas 3D.