26

Re: Let's see some of your scans

n2ri wrote:

"the poisson filter 13,7,1,1 settings seems to work the best for creating smooth surfaces."

where is this? in meshlab or Atlas settings? I just got my scanner and it came with the version 1.14 still and says no updates available when clicking to check. it dont seem to really check though its to fast snapping that response to actually check the site.

The the poisson filter is in Meshlab.  It is explained in Pirvan's tutorial.

27

Re: Let's see some of your scans

As a new user of FreeLSS, I find it encouraging that some people have so far managed to obtain what might be described as "usable" scans, sadly, this is not the case for me so far, but at least the slowly growing number of posted scans indicates that there is life in this project.

Many users report issues in the areas of calibration, lighting level, thresholds etc. that are not necessarily solved by the small number of posts available, as each test object is unique to the user who posted.  There is no consistency, no reference scan object that each and every user has access to.  By comparison, in the video imaging field for instance there are standard test charts available, allowing direct performance comparison to be performed.

The purpose of this post therefore is to make a plea for a standardised test object to be defined and made available as a printed object on Thingiverse.  An object of this type would need specific parameters to be defined, in order to provide "stress tests" for performance  parameters such as:
resolution on surface texture
dimensional accuracy
rotational consistency
tolerance to surface reflectivity
colour fidelity

There are no doubt many more which could be proposed. 
One existing popular model is the "Stanford Bunny" http://www.cc.gatech.edu/~turk/bunny/bunny.html, but this seems to lack suitable dimensional features.

Any ideas?

Reprap Prusa x1 ,  Reprap G frame Prusa x1,  OpenBuilds OX CNC x1, FreeLss/Ciclops 3D scanner x1

28

Re: Let's see some of your scans

You are absolutely correct in that your scan results are affected by all the aforementioned variables, primarily laser calibration and lighting.

Unfortunately, the reality is that achieving the kind of repeatable accuracy you're asking for, is going to be hard to do, because of the many variables inherent in the design.

It's a DIY scanner, so there can be lots of small variances from build to build that will account for the differences in resulting scans.

This is a tabletop scanner, which means the lighting source will be different from user to user.  This alone will account for 75% of all the discrepancies you see in reflectivity, color accuracy, and completeness of surface (missing parts)

The laser calibration, or lack of it, will account for the dimensional accuracy.

Camera focus will account for the surface resolution and accuracy.

Last but not least, laser scanners such as the Atlas have problems seeing certain areas, blind spots, and will require multiple scans from different angles to get most of the surfaces scanned.  Changing the scan angle will also affect the lighting (as seen by the camera), which means that even if your resulting point clouds are accurate, the color map (texture) will be different from model to model.

You could control the lighting by buying a photo box or tent with built in lighting, which would definitely help, but it's still no perfect solution.

As for a standardized model, the bunny would be a good idea, IF we could all have access to the same object.  The scanned version on the provided links is pretty devoid of fine details or color information so it wouldn't help much here.  Additionally, printing a 3D model for the purpose of scanning it, will yield different results due to the differences in printer capabilities.

Maybe if we could figure out some common household object that everyone has.  Or better yet, some common, inexpensive  that can be purchased in a store, like an action figure or something like it.

To print or, 3D print, that is the question...
SD3 printer w/too many mods,  Printrbot Simple Maker Ed.,  FormLabs Form 1+
AnyCubic Photon, Shining 3D EinScan-S & Atlas 3D scanners...
...and too much time on my hands.

29

Re: Let's see some of your scans

camera focus is very needed since none come set for the 'Macro' use of the scanner/s depending on which enclosure/case is used.

for me and some others purposed use. a Hot Wheels or Matchbox car would be a good object to test for reflection and details. also I think we may need (for such scans) a section here for best thing to use for dulling the shine on such objects like Talc or makeup powders in compacs etc. Talc free powders with no reflective materials in them and what color works best etc.

I have tried a few. the reason for this type powder is it can be easily wiped or blown off so it dont ruin model like paints etc and still leave details scanable.

Solidoodle 2 with Deluxe kit cover & glass bed with heater. and 2nd board SD2 used not 3rd and alum platform not installed yet still wood. also need cooling fan installed to board. use Repetier Host couple vers. Slic3r also have all free ware STL programs

30

Re: Let's see some of your scans

I just finished scanning another Warren Stratford figurine and posted a few images of it in the 3D scanning & Modelling forum. 

http://www.soliforum.com/topic/14364/th … s-3d-scan/

To print or, 3D print, that is the question...
SD3 printer w/too many mods,  Printrbot Simple Maker Ed.,  FormLabs Form 1+
AnyCubic Photon, Shining 3D EinScan-S & Atlas 3D scanners...
...and too much time on my hands.

31

Re: Let's see some of your scans

Pirvan, thanks for your reply.  I suppose there are at least two use cases for a desktop scanner - those that are primarily interested in obtaining aesthetically good scans (typically of ornaments, models etc), and those that are looking for a tool to reverse-engineer objects to a reasonable degree of dimensional accuracy, perhaps at the expense of other parameters such as colour fidelity.  For my sins, I fall into the latter category.

I gave a little thought to the construction of a test object which could be reproduced by anyone more or less anywhere, without spending too much time or money on it, as it has no other value than to allow users to compare results and hopefully gradually improve the scanning and especially post-processing workflow in a collaborative way. 

It is based on the universally available "coke" can as a well-defined structure for the attachment of reference items with known dimensions.  Below I have listed them from smallest to largest:

5 layers of masking tape up to 0.4mm
1 square of 0.1" prototyping board with 1mm dia holes
5 layers of P120 grit sandpaper up to up to 2.5mm
5 layers of P60 grit sandpaper up to 3.9mm
50mm lengths of M4 (0.7pitch), M6(1mm pitch), M8(1.2mm pitch) studding
50mm length of 0.1" pitch electrical header, 0.7mm pins
1 piece of lego brick 8 x 2 studs
4 nuts, M4, M5, M6, M8

Items were simply stuck on with hot melt glue and given a couple of coats of white primer spray.  The can can be placed either vertically or horizontally on the table to get H,V resolution. Here are a couple of views of one of the first scans at 1.2Mpixel resolution.

http://soliforum.com/i/?zUwn9It.pnghttp://soliforum.com/i/?JgxvUpv.png

One of the more obvious limitations of the scanner is the absence of detail in shadows caused by scanning in an anti-clockwise direction only.  Would it not be possible to repeat the scan in a clockwise direction and merge the result?

If anyone is interested in constructing the 5C calibrator,  I would be happy to post the STL file.

Post's attachments

5C100.png
5C100.png 161.72 kb, file has never been downloaded. 

5C201.png
5C201.png 157.67 kb, file has never been downloaded. 

You don't have the permssions to download the attachments of this post.
Reprap Prusa x1 ,  Reprap G frame Prusa x1,  OpenBuilds OX CNC x1, FreeLss/Ciclops 3D scanner x1

32

Re: Let's see some of your scans

I have definitely had some objects where being able to do a scan in a clockwise direction would have captured spots that couldn't be seen by the scanner with a counter clockwise scan no matter where I put it on the table. I'm not sure if it's feasible or not but an option to do a clockwise scan would be a very useful enhancement.

Bob

33 (edited by pirvan 2016-03-28 18:08:22)

Re: Let's see some of your scans

I'm not sure I follow you guys. 

Why would the rotation of the turntable during a scan make any difference.  If you scan with both lasers, under the same lighting conditions, the 1600 individual slices it photographs (assuming you do 1600 steps/rotation), are the same.  The order is reversed, but the images are identical.

You can prove it to yourself by turning on the lasers, and looking at the camera view, while rotating the turntable by hand in either direction.  Given a specific location/angle you stop at, there's no difference which direction you came from.  The picture the camera sees is the same.

Regarding shadows.  If a particular area is in a shadow cast by an object or another part of the model itself, it shouldn't matter, because as long as the laser line can shine on it it will pick it up.  There will be no color information or the colors may be dark, but the surface will be seen by the laser.

On the other hand, if the shadow you speak of, is actually an obstruction that prevents the laser from touching it, then that will be an area without any data (a hole).  For example you can have a a model of a person with their arm stretched in front or side. The laser will hit the arm, but will not see the area of the body behind the arm.  Those are the blind spots I've referred to in the past and in other posts.

Turntable laser scanning is fraught with such problems.  Unless you're scanning a more or less cylindrical object, there will be many of these "dead" blind spots.

This is the reason why I do more than one scan.  I scan from multiple angles, in an attempt to capture as much data as possible.

If you're scanning an object which has many depressions that create blind spots, one way to get around it is to bring the object closer to the front of the turn table (about 2" from the center), do a scan that way, then while still facing the same way, move the model toward the back of the turntable the same amount. 

The change in angle will definitely help see more of the blind areas.

To print or, 3D print, that is the question...
SD3 printer w/too many mods,  Printrbot Simple Maker Ed.,  FormLabs Form 1+
AnyCubic Photon, Shining 3D EinScan-S & Atlas 3D scanners...
...and too much time on my hands.

34

Re: Let's see some of your scans

I do turn both lasers on and rotate the turntable by hand to try and identify where on the table will allow the lasers to be best visible to the camera prior to each scan.  I find this especially useful when trying to figure out where to put the object to get the best picture of the hardest to capture areas.

What you say makes perfect sense, however for whatever reason, be it user error, lighting or some other factor I believe in some cases that the laser hit the object better/worse in the same spot depending on table rotation.  Again this is only me manually rotating the table and looking at the laser so in reality it's entirely possible that the scan results would have been identical irregardless of table direction.

Bob

35 (edited by pirvan 2016-03-30 22:23:23)

Re: Let's see some of your scans

I think that might have some (minor) impact if you were to capture in video mode. 

In still mode, the turntable rotates, stops, then the camera takes 2 pictures for each laser.  One with the laser off, one with the laser on.  The "laser on" image is like the image you see when you do a test, black with only the red laser trace.   FreeLSS uses this image as a mask against the image taken with the laser off, saving only the little band of color that was under the laser.  Then it rotates 0.225° (or whatever depending on your number of steps), and repeats the steps above.

At least that's how I understand it works.

To print or, 3D print, that is the question...
SD3 printer w/too many mods,  Printrbot Simple Maker Ed.,  FormLabs Form 1+
AnyCubic Photon, Shining 3D EinScan-S & Atlas 3D scanners...
...and too much time on my hands.

36

Re: Let's see some of your scans

That's correct.  The direction the table turns shouldn't matter.

37

Re: Let's see some of your scans

Thanks to all for comments relating to the direction of rotation, it has certainly helped me towards a better understanding of what is involved in obtaining good scans. I have since looked in greater detail at the "shadowing" in my previous scans which gave the impression of a preferential bias towards objects rotating away (CCW) from the camera over those approaching (CW).  In the case of this test object it is co-incidental that the missing detail on faces on the RHS side were due to the faces being tangential to the can and therefore more closely parallel with the laser planes than the LHS. It tends to confirm the process of moving the object so that both close and distant views are obtained of the same detail, as this would increase the number of points per tangential face.  This led me to wonder if deliberately offsetting one laser plane  by say 2 degrees with respect to the other (assuming the software could take this into account) would yield a more beneficial effect than the existing equi-angled arrangement.


Taking the point about direction not being a factor, it raises a few questions though.  Assuming first that the lasers are setup identically and creating the same imaging planes equi-centred on the imager.  I have not yet looked into the detail of the processing, but is it fair to assume that there is no "master" laser plane - that is both have equal significance when the software automatically merges them?  Is is better to create two seperate files and work on each L,R image seperately?

Thanks again,

Reprap Prusa x1 ,  Reprap G frame Prusa x1,  OpenBuilds OX CNC x1, FreeLss/Ciclops 3D scanner x1

38

Re: Let's see some of your scans

I'm not sure, and this is something that Uriah needs to comment on, but if I remember correctly, the final "blend" is made up primarily of the right hand laser, with the left laser filling in the missing spots.  In other words, when the same information exists from both lasers, only the right laser info is kept.

To print or, 3D print, that is the question...
SD3 printer w/too many mods,  Printrbot Simple Maker Ed.,  FormLabs Form 1+
AnyCubic Photon, Shining 3D EinScan-S & Atlas 3D scanners...
...and too much time on my hands.

39

Re: Let's see some of your scans

When scanning with both lasers, the points from the left laser are only used where no points exists for the right laser.

- Uriah

40 (edited by knowack 2016-05-04 10:59:43)

Re: Let's see some of your scans

Here's my first scan with Atlas 3D before any clean-up or post processing (a screen grab of the browser view):

https://www.dropbox.com/s/rogea5nejl5rfh8/bunny.JPG?dl=1

This is a ceramic bunny about 5" tall, that belonged to my grandmother.  It had been dropped and glued back together at some point.  I intend to 'repair' the glue joints and broken paw, and print a new one.  I'm quite impressed that the Atlas 3D even picked up the wicker texture of the basket!

I used the 'Default' preset, with the exception of using both lasers.

I have to say that I am very pleased with the Atlas 3d scanner, and laser-cut acrylic case!  Kudos to Uriah and Tim for such a great piece of hardware!

I was a Kickstarter backer, but only recently found time to put the Atlas 3D together.  Although I had originally intended to print the scanner (mostly for bragging rights), I'm glad I bought the laser-cut case.  It looks great, and I probably would have wasted more $ in filament (I'm very picky about my prints) than the cost of the acrylic frame.  I'm confident my results will only improve as I play around with different settings.

My next project for the Atlas 3D is to get WebGL working in the Pi's browser!

-Kevin

41

Re: Let's see some of your scans

Another screen grab of an Atlas 3D scan:

https://www.dropbox.com/s/68ydf2rsohk1del/bird.JPG?dl=1

This is a 4" tall carved and painted wooden bird that I bought in Panama several years ago.  The interesting thing is that it has a glossy finish, yet the scanner wasn't adversely affected; this same item caused difficulty for a Matter&Form scanner.

42

Re: Let's see some of your scans

what lighting are you using?

Solidoodle 2 with Deluxe kit cover & glass bed with heater. and 2nd board SD2 used not 3rd and alum platform not installed yet still wood. also need cooling fan installed to board. use Repetier Host couple vers. Slic3r also have all free ware STL programs

43

Re: Let's see some of your scans

I'm using a fluorescent bulb at the moment, but it is a little too blue-ish.  I think I'm going to go with halogens.

-Kevin

44

Re: Let's see some of your scans

A rubber ducky! I ended up scanning it in dim lighting to get the most accurate results.  it did a pretty good job picking up the contours of the wing, etc.  A hi-res scan using only the right laser.

https://www.dropbox.com/s/o4ktxjowddvxn4j/Ducky.JPG?dl=1

-Kevin

45

Re: Let's see some of your scans

I have a very similar duck that I sometimes scan as well smile

46

Re: Let's see some of your scans

Here are some pics of my very first scan, of a fish tank ornament. 5 MP @ 800 steps. I'm going to try again at laser/camera calibration and fiddle with the settings, etc.

I assume the lines are caused by laser alignment? I was able to refocus the camera, seems pretty sharp.

http://soliforum.com/i/?Iv5xh2U.pnghttp://soliforum.com/i/?8bRxMJ8.pnghttp://soliforum.com/i/?eDQYugH.png

47

Re: Let's see some of your scans

Here's a new scan I just did, 1600 frames this time, 5 MP still, both lasers. I re-calibrated before this scan as well, and took a lot of time to get things just as perfect as possible. Also, updated to the latest version of freelss.

Also I put the scanner in a lightbox with an LED hue light so I could adjust the intensity and color temperature. I noticed that the lasers have some difficulty getting into the fine details of the head. I've ordered higher quality and higher power line lasers off ebay so I'll experiment with that, as well as the higher resolution pi camera.

Overall I'm very impressed so far, and while there is room for improvement the performance is already quite good with some initial fiddling.

http://soliforum.com/i/?CHVkeH3.pnghttp://soliforum.com/i/?yo3XkEe.png

48 (edited by czc 2016-07-12 15:07:37)

Re: Let's see some of your scans

One of my first attempts using FreeLSS 1.18.
8MP v2 camera, RPI3

Scanned in 8MP, 200 frames, both lasers. Very fast just to play with lighting and so on.
No postprocessing as it is captured directly from the scanner's webpage.

Model is 5cm tall.

Amazing software!

Post's attachments

1.png 67.34 kb, file has never been downloaded. 

2.png 98.58 kb, 2 downloads since 2016-07-12 

You don't have the permssions to download the attachments of this post.

49 (edited by C_D 2016-08-06 22:07:42)

Re: Let's see some of your scans

Just fitted some new lasers and recalibrated my scanner. Pretty happy so far.

This is 1100 frames (thats full stepping using my 0.9deg stepper and 1:2.75 geared platform). Scanned with the left and right lasers separately and then merged in meshlab. Minimal post processing, just removing a few clusters of stray points caused by shiny spots on the part. BIG thanks to pirvan for the post processing tutorials!

There are some scaling issues I need to work out, the scan is about 50% scale and stretched slightly in Z. I scaled the STL down by 80% in the Z axis only and it now looks about right.

http://soliforum.com/i/?WIriu3x.jpg

http://soliforum.com/i/?zN6XvJI.jpg