First scans with lowbudget rig

What have you scanned recently?
Tzk
Posts: 23
Joined: 17 Mar 2020, 13:28

First scans with lowbudget rig

Post by Tzk » 03 Aug 2020, 08:25

Hi all,
thought i'd share the first steps with my sls rig.

Setup:
Viewsonic PLED-W500 (1280x800p DLP LED diamond)
YW500U3M camera 5mpix USB3 YUY2
ZLKC 12mm F1.4 C-mount lens
DIY 150mm panels made from cardboard and 20mm wood

Total hardware costs so far: ~300€

The first calibration gave me a RMS value of 0.189:

Code: Select all

 - RMS error is ok: 0.189753
Camera 0 [YW500U3M (2FEC032D)] calibrated:
<camera_model>CameraModelTsai</camera_model>
<cx>1352.6744308236191046</cx>
<cy>841.39479384338369528</cy>
<f>6672.2796385739211473</f>
<sx>1</sx>
<kappa1>-1.1355731441117723929e-13</kappa1>
<resX>2592</resX>
<resY>1944</resY>
<nx>0.50383805549141680302</nx>
<ny>-0.01948580120589021164</ny>
<nz>-0.86357832151461078052</nz>
<ox>0.25183835582838937084</ox>
<oy>-0.95299922754555432824</oy>
<oz>0.16843371049534830441</oz>
<ax>-0.82627153912759321308</ax>
<ay>-0.30234545779446025948</ay>
<az>-0.47525000555368396293</az>
<px>655.71913985958099147</px>
<py>241.0697129790300437</py>
<pz>441.03874242193506916</pz>

Horizontal open angle: 21.9432
Vertical open angle: 16.6166
Calibrate Projector.
~
~
~
CalibrateProjector: leftBackPlane  = (p=5.36432e-17 5.36432e-17 57.1261 1 n=1 -0 -6.12323e-17 0)
CalibrateProjector: rightBackPlane = (p=55.2053 0.314087 -1.27935e-16 1 n=0.00618138 1.59425e-05 0.999981 0)
Angle between calibration panels: 90.3542 degree
  Size of camera's center pixel at background plane: 0.201245 mm
cy=-0.5 rmsError=0.00409419
cy=0    rmsError=0.00409419
cy=0.5  rmsError=0.000358095
cy=1    rmsError=0.000358095
cy=1.5  rmsError=0.000358094
Then i tried to scan a small rabbit statue (20cm high) and here's the result of 18 scans combined. I used a resolution of 2000 and sharpness 1. So far i'm pleased with the result. Next up i'll try to scan with my laptop as this was made on my desktop machine and i want a portable scanner.
hase-cut.jpg
And for comparison the same rabbit, but scanned with photogrammetry (Sony Nex 5T, 1650SELP). The SLS quality is decent and i'm very happy so far.
hase-photogrammetrie.jpg

User avatar
OBNRacerMan
Posts: 173
Joined: 12 Aug 2019, 09:46
Location: Obninsk, Russia

Re: First scans with lowbudget rig

Post by OBNRacerMan » 03 Aug 2020, 09:19

RMS value of 0.189 - this is an excellent result (I usually have RMS 1.2 - 0.9 at full resolution)

Where photogrammetry worked better visually - these are the eyes (they are round), but they are not on the scan

Tzk
Posts: 23
Joined: 17 Mar 2020, 13:28

Re: First scans with lowbudget rig

Post by Tzk » 03 Aug 2020, 09:55

I forgot to mention the eyes, sorry for that. The eyes on the photogrammetry scan were edited in afterwards with Meshmixer. These are glossy and transparent and thus i couldn't scan them with either technique. Besides that both scans are completely ontouched and directly out of cam (or out of scanner) :D

I've seen RMS values of 0.13 to 0.22 so far, depending on my patience while calibrating. I guess the RMS will become bigger when i use bigger panels? I currently got a sheet of DIN A3 paper which results in ~150mm panels.

User avatar
Micr0
Posts: 426
Joined: 15 Nov 2016, 15:20

Re: First scans with lowbudget rig

Post by Micr0 » 03 Aug 2020, 12:14

Tzk wrote:
03 Aug 2020, 09:55
I forgot to mention the eyes, sorry for that. The eyes on the photogrammetry scan were edited in afterwards with Meshmixer. These are glossy and transparent and thus i couldn't scan them with either technique. Besides that both scans are completely ontouched and directly out of cam (or out of scanner) :D

I've seen RMS values of 0.13 to 0.22 so far, depending on my patience while calibrating. I guess the RMS will become bigger when i use bigger panels? I currently got a sheet of DIN A3 paper which results in ~150mm panels.
RMS is simply a way to find an average of a continuously changing function. Like a sine wave. It shouldn't be dependent on the size of your calibration panels.
µ

User avatar
OBNRacerMan
Posts: 173
Joined: 12 Aug 2019, 09:46
Location: Obninsk, Russia

Re: First scans with lowbudget rig

Post by OBNRacerMan » 03 Aug 2020, 16:21

Micr0 wrote:
03 Aug 2020, 12:14
RMS is simply a way to find an average of a continuously changing function. Like a sine wave. It shouldn't be dependent on the size of your calibration panels.
My RMS value is more affected by the resolution of cameras (if you make a low resolution, then the result will be better on the same table) I also noticed that if the table is smaller than the area indicated in the center of the program screen, then the edges of the scan, even with a good RMS bend strongly
dualcam_SUA_01as.jpg
Full resolution (6.3Mp) - one edge of the table is slightly off the zone
dualcam_SUA_01bs.jpg
Half resolution (FullHD) - the table completely occupies the configuration area

hoffybeefe
Posts: 8
Joined: 07 Aug 2019, 13:03
Location: QLD, Australia

Re: First scans with lowbudget rig

Post by hoffybeefe » 04 Aug 2020, 02:25

OBNRacerMan wrote:
03 Aug 2020, 16:21
RMS value is more affected by the resolution of cameras
aah HUH! this explains why, with my new MV Camera, that my RMS is really high compared to how it used to be using my Canon HF G30. I expected my RMS value to be much lower (or at least similar) with the better quality MV Camera and i was shocked when it was seemingly worse (but the scan quality itself is better).

Thanks for the accidental tip OBNRacerMan. :)

Tzk
Posts: 23
Joined: 17 Mar 2020, 13:28

Re: First scans with lowbudget rig

Post by Tzk » 04 Aug 2020, 09:22

Half resolution (FullHD) - the table completely occupies the configuration area
By coincidence i did it like shown in the second image. Lucky me i guess... I also haven't had issues with bend scans, however i'm using the calibration angle compensation (i got 90.3° with my plates).

Here's another test scan i did yesterday. The object is about 10cm tall, still the same 150mm calibration plates were used. I had to coat the object with dryshampoo and i guess that's why the surface is rather bumpy. I made two full rotations with 18 scans each (36 total).

Also switching the refresh rate of my projector to 50Hz reduced the surface noise a bit. I used 1/64 exposure for the camera and 50fps. Seems to work well in this combination.
2.jpg

kth307000
Posts: 19
Joined: 17 Jul 2020, 15:35

Re: First scans with lowbudget rig

Post by kth307000 » 04 Aug 2020, 11:03

Tzk wrote:
03 Aug 2020, 08:25
Hi all,
thought i'd share the first steps with my sls rig.

Setup:
Viewsonic PLED-W500 (1280x800p DLP LED diamond)
YW500U3M camera 5mpix USB3 YUY2
ZLKC 12mm F1.4 C-mount lens
DIY 150mm panels made from cardboard and 20mm wood

Total hardware costs so far: ~300€

The first calibration gave me a RMS value of 0.189:

Code: Select all

 - RMS error is ok: 0.189753
Camera 0 [YW500U3M (2FEC032D)] calibrated:
<camera_model>CameraModelTsai</camera_model>
<cx>1352.6744308236191046</cx>
<cy>841.39479384338369528</cy>
<f>6672.2796385739211473</f>
<sx>1</sx>
<kappa1>-1.1355731441117723929e-13</kappa1>
<resX>2592</resX>
<resY>1944</resY>
<nx>0.50383805549141680302</nx>
<ny>-0.01948580120589021164</ny>
<nz>-0.86357832151461078052</nz>
<ox>0.25183835582838937084</ox>
<oy>-0.95299922754555432824</oy>
<oz>0.16843371049534830441</oz>
<ax>-0.82627153912759321308</ax>
<ay>-0.30234545779446025948</ay>
<az>-0.47525000555368396293</az>
<px>655.71913985958099147</px>
<py>241.0697129790300437</py>
<pz>441.03874242193506916</pz>

Horizontal open angle: 21.9432
Vertical open angle: 16.6166
Calibrate Projector.
~
~
~
CalibrateProjector: leftBackPlane  = (p=5.36432e-17 5.36432e-17 57.1261 1 n=1 -0 -6.12323e-17 0)
CalibrateProjector: rightBackPlane = (p=55.2053 0.314087 -1.27935e-16 1 n=0.00618138 1.59425e-05 0.999981 0)
Angle between calibration panels: 90.3542 degree
  Size of camera's center pixel at background plane: 0.201245 mm
cy=-0.5 rmsError=0.00409419
cy=0    rmsError=0.00409419
cy=0.5  rmsError=0.000358095
cy=1    rmsError=0.000358095
cy=1.5  rmsError=0.000358094
Then i tried to scan a small rabbit statue (20cm high) and here's the result of 18 scans combined. I used a resolution of 2000 and sharpness 1. So far i'm pleased with the result. Next up i'll try to scan with my laptop as this was made on my desktop machine and i want a portable scanner.

hase-cut.jpg

And for comparison the same rabbit, but scanned with photogrammetry (Sony Nex 5T, 1650SELP). The SLS quality is decent and i'm very happy so far.

hase-photogrammetrie.jpg
This is a very nice result.

I am using the same camera as you and the project is different.
But the result is different.

When scanning at the maximum resolution of the camera, except for the scan quality, the computer slows down and the operation is not smooth.
So I scanned the shoes with 2048X1536.

I received a scan with a total of 32 cuts and the quality is not satisfactory.
The data received in 32 cuts is about 1.8 gigabytes.

I think the file is too big even though I did not apply the maximum resolution.

Another question is what is RMS above.
And where do you get that data from?

Everything is unfamiliar because there is no knowledge of the camera and scan.

Tzk
Posts: 23
Joined: 17 Mar 2020, 13:28

Re: First scans with lowbudget rig

Post by Tzk » 04 Aug 2020, 14:24

kth307000 wrote:
04 Aug 2020, 11:03
I received a scan with a total of 32 cuts and the quality is not satisfactory.
The data received in 32 cuts is about 1.8 gigabytes.

Another question is what is RMS above.
And where do you get that data from?

Everything is unfamiliar because there is no knowledge of the camera and scan.
Regarding the filesize: i had >4gb with around 40 scans. I cut the background which reduces the filesize and then aligned the scans afterwards. This does also save processing time.

And RMS:
Open the advanced settings menu (top right corner), go to the "service" submenu and tick the box for "showdebugconsole". A new console window should appear. Now calibrate your setup and have a look at the console window. Scroll a bit upwards and you'll find the output i posted above including the RMS value and even the angle between your calibration panels.

I don't have issues with slowdowns, however i'm using a Intel 7700K and 16gb of ram. I can try on my old trusty Dell XPS (Core i5 460m, about 10 years old) to see if it's an issue with processing power.

Tzk
Posts: 23
Joined: 17 Mar 2020, 13:28

Re: First scans with lowbudget rig

Post by Tzk » 27 Aug 2020, 11:22

OBNRacerMan wrote:
03 Aug 2020, 16:21
My RMS value is more affected by the resolution of cameras (if you make a low resolution, then the result will be better on the same table) I also noticed that if the table is smaller than the area indicated in the center of the program screen, then the edges of the scan, even with a good RMS bend strongly

I noticed that a) my scans bend and b) RMS gets insanely bad when i'm too far away from the panel. Here's two calibrations with RMS 0.2 and 0.4 for comparison. I only moved the panels about 20cm backwards and adjusted the focus for the 2nd calibration.

I guess that's how i got RMS 0.13 when i first tried the setup... Anyways, i'll keep an eye on it in the future.
Capture.JPG
Capture2.JPG
I also practiced a bit. Here's my trusty Makita drill. I had to take ~70 scans and threw 30 away (so 40 used for the fused result). I guess that are too many scans, right? I also only scanned one half of the drill...
Capture3.JPG

Post Reply