Getting good scans (Page 1 of 3)

Update 1

I did another scan, this time with the laser threshold at 3.  It looks like it picked up more of the model, especially on the feet and the back, but it also started to pick up extraneous stuff that's not really there.

Anyway, I also brought the PLY file into MeshLab and started playing with the various surface reconstruction filters.  Here are the results using the Poisson Reconstruction filter.  You can see where the filter attempted to guess how to fill the missing parts.

Couple of intresting notes:

I'm not sure the camera can be focused, I think it's a fixed focus device.  It appears its minimum focal distance is about 4-5ft., certainly not 10". 

This is something that Uriah needs to address, unless I just got the only weird one of the batch.

I did find a way around this (sort of), I put a pair of reading glasses with 2.5x magnification in front of the camera, and it brought things in better focus.  I think something closer to 3-3.5x would do the trick.  But it's certainly a kludge.  It really needs to be addressed properly.

The magnification still didn't fix the basic scanning problems, but it has improved the camera's ability to see finer details.

pirvan,  if you could help me I would be greatly thankful.

I was getting decent scans until today .   I did the update and all went bad.   After recalibrating several times and pulling most of my hair out I realized ( I think ) that some of my settings might have changed. If you could check my settings against yours it might help .

My settings  then setup

camera mode 1.9 video
laser threshold 2
cam y  3.5
cam z  10.25
right laser x  4.95
right laser y  3.5
right laser z  10.25
left laser x  -4.95
left laser y  3.5
left laser z  10.25

I know I will have to tweak them for my scanner but at least it will give me a starting point.

Thanx  Bill D

Thanks   I guess it back to the drawing board.

Thanks for this.  I'll do the surgery in the next day or so.

Meanwhile, I'm really struggling with getting usable scans from this thing.  I have played around a lot with lighting, and the laser threshold, I've downloaded the instruction manual from the matterandform scanner which uses similar technology.

But unless  I scan something simple and basic like a box or a bottle, I get really weird artifacts, incomplete areas, distorted areas, and when I scan with both lasers, I get what appear two be to shells, the main one, and a larger one, sort of encompassing part of the main body.

Here are a couple of scans.  Notice the first one, and how it's pinched in the middle, as if it twisted itself around the center.  The one on the right is missing lots of areas, which I know are seen by the laser and the camera.

Edit:
Another thing I'm noticing is that since I upgraded to v.1.5 of the FreeLSS software, it now takes a lot longer to do a scan, and seems to be primarily due to the fact that the scanner now takes 2 or more snapshots per step.  Whether it actually does 2 snapshots or just flashes the laser twice, I'm not sure, but it appears to do two or (sometimes) more laser flashes depending on something (not getting a good read maybe?).

Is this normal?

pirvan, were the objects that you scanned shiny? I only see a photo of one and that seems a bit shiny.

MOst of these objects are small toys made of plastic or rubber.  Most of them are satin finish, but they might reflect in some places (usually the brighter colors).

If the scanner is this sensitive to reflections, what can I do to reduce it.

Edit:  This assumption was wrong.  See post below.

Yes, if I scan with a single laser, there's no "shell".  Only one , but with loads of missing pieces.

I think I figured out why the shell occurs, and it has to do with the angle of incidence, basically perspective. 
Each laser see the object from a different angle.  The further out the model sticks out from the center of the turntable the, more pronounced the perspective.  The closer the the surface is the the center of the turntable (the point where the to beams converge, the less the perspective distortion, and the closer the 2shells are.  At the center of the turntable the two shells would merge.

So a cylindrical model would not show this problem, but add a couple of "arms" sticking straight out of it, and you instantly have a problem.  The software doesn't seem to account for this.

Unless I'm wrong, of course.  I just wish someone would enlighten me.

Which video mode are you using?

Did you check your calibration?  I calibrate mine with a wood maker's square to ensure everything is square.

I tried all the modes, including default, hi-res video and hi-res still.

For calibration I used the calibration thingy that Uriah made, and it's square.  Both beams converge in the middle of the turntable and appear as a single beam on the vertical surface of the calibration thingy.

However, things being out of square doesn't explain the discrepancy in size between the 2 lasers.

You must remember that you are never going to get a perfect scan by:
1.  Doing one scan.  You must do multiple scans from multiple angles and then combine the scans in Meshlab or something.
2.  Do no post scan processing.

I did measure everything, and also adjusted for the actual elevation of the camera and lasers.  Given that the software links some of the dimensions, it assumes that everything is symmetrical, and my measurement bear that, within 1 mm or so.

Calibrations done as per instructions are correct, and the 2 lasers converge into a single vertical beam at the center of the turntable.

Last night I did a scan at hi-res, and 3200 steps, which really improved the quality of the scan, however it took over 2 and 1/2 hours to complete the scan from 1 (left) laser only ! 

The problem with the single laser is that there were still many areas the laser couldn't see, that the right laser would have, but since the 2 lasers produce  different size scans, I can't simply blend them.

So I need to understand why this discrepancy exists.  I tried playing with the one setting that can be independently adjusted, the Left Laser X dimension, which I changed from -4.5" to -5.5".  And while I did see some small changes, they were mostly related to the angle of the scan and the perspective correction.  It didn't really fix the size of the resulting scan.

I know that some people had problems with this on Google+ groups, but I haven't seen any resolutions.  Besides, it appears the Google+ groups are pretty much dead, since Uriah stated he'll concentrate his support efforts on this forum.

I'd really like to get to the bottom of this problem.  At this stage, I'm pretty sure it's software related.