Notifications
Clear all

[Solved] Two laser troubles  

Page 1 / 3

jwalt
Posts: 6
(@jwalt)
Active Member
Joined: 9 months ago

Hi,

Today I added a second laser to my Fabscan Pi, and I am getting wrong results. The scan from the first laser is about 10% smaller in the X and Y dimension than the scan from the second laser. The error seems to be less pronounced in the Z direction. Both point clouds are also rotated relative to each other (about 1-2 degrees).

After making sure that the lasers cross the exact center of the turn table, the camera also points exactly at the center (both with less than 1mm offset), and making sure the lasers are exactly vertical, and recalibrating, I am at a loss what else I can try.

Additionally, for this model I always get points in thin air, always under the arm. I covered the model in blue spray chalk to eliminate any reflection and reduce indirect illumination, but it didn't help.

Many other parts look pretty decent, but due to these issues, the scan is ultimately unusable. I was unable to fix this up in meshlab.

It's really frustrating to have bought this device and not being able to produce a single successful scan in three years. Please help.

Reply
17 Replies
mario
Posts: 451
Admin
(@mario)
Member
Joined: 2 years ago
Posted by: @jwalt

Today I added a second laser to my Fabscan Pi, and I am getting wrong results. The scan from the first laser is about 10% smaller in the X and Y dimension than the scan from the second laser. The error seems to be less pronounced in the Z direction. Both point clouds are also rotated relative to each other (about 1-2 degrees).

When the point clouds do not match each other, the hardware setup is not aligned correctly. Some important points are: 

  • check that the checkerboard sheet is in the vertical center of the calibration stand.
  • the vertical green alignment line of the laser needs to hit the center of the turntable
  • both lasers need to cross the center of the turntable
  • both lasers should be vertical aligned on the calibration pattern
  • the camera should be as sharp as possible (focus alignment)

You need to recalibrate if you change the physical setup.

Here are some helpful hints: https://fabscanpi-server.readthedocs.io/en/latest/scanner_calibration.html#laser-alignment

The calibration part is the hardest software part of this project. I am still trying to improve it. All i can say, it is possible to reach good results. Have a look at https://fabscan.org/community/general/new-software-release-v-0-8-1/ @Danilo0702 made some nice looking scans. Maybe he can share his experience. 

I know that it can be frustrating when the pointclouds do not align 100%. But that is the reason why i added an option which saved a single cloud for each laser. This gives you the opportunity to merge the cloud by using CloudCompare or Meshlab.

Posted by: @jwalt

Additionally, for this model I always get points in thin air, always under the arm. I covered the model in blue spray chalk to eliminate any reflection and reduce indirect illumination, but it didn't help.

Have you tired to use different threshold settings. Can you post some pictures of the object and your scan results please. That will maybe help me to understand what affects those points.

I tried a lot of different 3D Scanners in the last years, and out of my experience i can say that there was not a single one which was able to put the object into the scanner and getting a perfect result. Even very expensive ones. In most cases some rework on the pointcloud was required (cleaning/editing etc.)

Reply
1 Reply
jwalt
(@jwalt)
Joined: 9 months ago

Active Member
Posts: 6
Posted by: @mario

When the point clouds do not match each other, the hardware setup is not aligned correctly. Some important points are: 

  • check that the checkerboard sheet is in the vertical center of the calibration stand.
  • the vertical green alignment line of the laser needs to hit the center of the turntable
  • both lasers need to cross the center of the turntable
  • both lasers should be vertical aligned on the calibration pattern
  • the camera should be as sharp as possible (focus alignment)

You need to recalibrate if you change the physical setup.

Yes, that's what I meant. I double checked that every single point you mention is true. Any remaining deviation was way below 1mm, not noticeable to the naked eye. A second calibration run reduced the error to about 5% instead of 10%. The end result is still unusable, because there is the slight rotation offset and a tiny shift to the side as well. The latter two errors are worse than the scaling, because it's really difficult to fiddle with 5 unknown values at the same time. Do you have any ideas for that?

Looking at the results of @Danilo702 I wonder what is going wrong. There must be some other factor that also influences calibration. But it's good to see that the FabScan is able to produce good results. I need to test more...

Regarding my model-specific troubles, yes, I have screenshots, but I don't see an option to post images here?

 

Reply
mario
Posts: 451
Admin
(@mario)
Member
Joined: 2 years ago
Posted by: @jwalt

Yes, that's what I meant. I double checked that every single point you mention is true. Any remaining deviation was way below 1mm, not noticeable to the naked eye. A second calibration run reduced the error to about 5% instead of 10%. The end result is still unusable, because there is the slight rotation offset and a tiny shift to the side as well. The latter two errors are worse than the scaling, because it's really difficult to fiddle with 5 unknown values at the same time. Do you have any ideas for that?

I am working on some solutions, for a software error correction. But i can't say when this is finished. 

Posted by: @jwalt

Looking at the results of @Danilo702 I wonder what is going wrong. There must be some other factor that also influences calibration. But it's good to see that the FabScan is able to produce good results. I need to test more...

Until that is done we need to find out what's going on ...

Posted by: @jwalt

Regarding my model-specific troubles, yes, I have screenshots, but I don't see an option to post images here?

 

I think that is the Forum spam prevention feature. I just realized that your account is a fresh one. I think after a couple of posts you should be able to see the file upload Link below the text field.  

Reply
jwalt
Posts: 6
(@jwalt)
Active Member
Joined: 9 months ago

Indeed, now I can attach files. I don't have a picture of the model available right now, but just look at the end of the right arm. It is supposed to be a free hanging arm. I have no idea where those few lines of dots hanging in mid air come from. They look like echos from the last correct line of the arm

artifacts

 

Reply
1 Reply
Danilo0702
(@danilo0702)
Joined: 2 years ago

Reputable Member
Posts: 294
Posted by: @jwalt

Indeed, now I can attach files. I don't have a picture of the model available right now, but just look at the end of the right arm. It is supposed to be a free hanging arm. I have no idea where those few lines of dots hanging in mid air come from. They look like echos from the last correct line of the arm

artifacts

 

i has det same problem.... try and calibrer the 2 lasers on exakt the center ---- verry importend is the temperature, since i use the coolingsystem is evrything ok

Reply
mario
Posts: 451
Admin
(@mario)
Member
Joined: 2 years ago
Posted by: @jwalt

Indeed, now I can attach files. I don't have a picture of the model available right now, but just look at the end of the right arm. It is supposed to be a free hanging arm. I have no idea where those few lines of dots hanging in mid air come from. They look like echos from the last correct line of the arm

That is strange. I will try to reproduce such an error. And this happens even when you used some spray on the object surface?

Reply
1 Reply
jwalt
(@jwalt)
Joined: 9 months ago

Active Member
Posts: 6

Yes, indeed. It did not change no matter what I did. These artifacts were there when I first tried this object. Then I bought white spray chalk, but no change. Finally I bought blue spray chalk, thinking that it could be indirect light, and blue won't reflect red light less. But no change either.

Actually, since the object is not terribly important to me, I could ignore this and hope this is a one-time problem. I could also send it to you so you can try yourself.

Reply
Page 1 / 3
Share: