Dear DAVID Community,

As you are probably aware, the entire DAVID company, team and product range has been integrated into HP Inc.
This forum is now read-only. Relevant content has been migrated to HP and merged into HP's Support Forums on November 1, 2016.

To start new discussions for 3D Scanning, please register and post your new topic at the HP Support Forums

Incorrect alignment when fusing scans

Problems, solutions and discussion about scanning with a video projector and stripe patterns.

Incorrect alignment when fusing scans

Postby vvs » Mon Mar 21, 2016 7:13 pm

Hi

I am not getting good results when aligning scans via shape fusion. Wondering what I am doing wrong? I have tried pair wise alignment, Y-axis alignment and global fine registration. In the shape fusion preview, the scans look like they are aligned. But when I fuse, the result is clearly not aligned. Greatly appreciate any help fixing this.

Individually the scans look great and the quality is outstanding.
Image

Here is another individual scan of the same model from a pose that is a few degrees rotated.
Image

However when I fuse these two poses the result is clearly misaligned. You can see this by the blurring on the arms and face
Image

Here is my camera calibration file:
Code: Select all
<camera_model>CameraTsai</camera_model>
<cx>501.92642682276949</cx>
<cy>383.14486273013495</cy>
<f>2494.0886316419051</f>
<sx>1.0040930299601418</sx>
<kappa1>-7.5660356191587861e-010</kappa1>
<resX>1056</resX>
<resY>704</resY>
<MotionType>FixedPose</MotionType>
<Pose>
<nx>0.72150965110799825</nx>
<ny>-0.0097041962957734214</ny>
<nz>-0.69233637195532927</nz>
<nw>0</nw>
<ox>0.13292819918658774</ox>
<oy>-0.97936100296171613</oy>
<oz>0.15225675597106555</oz>
<ow>0</ow>
<ax>-0.67952477307234782</ax>
<ay>-0.2018857460349166</ay>
<az>-0.70532845421753665</az>
<aw>0</aw>
<px>463.86932639564981</px>
<py>92.353472366822444</py>
<pz>477.50786661711527</pz>
<pw>1</pw>
</Pose>


Here is my projector calibration:
Code: Select all
<camera_model>CameraTsai</camera_model>
<cx>0.47956703297721159</cx>
<cy>1.0557649259390753</cy>
<f>2.4048681408632513</f>
<sx>0.75529764823315015</sx>
<kappa1>0.0072226815754539925</kappa1>
<nx>0.50142183176690325</nx>
<ny>-0.0037616114188919747</ny>
<nz>-0.86519477397118882</nz>
<ox>0.27639578264144804</ox>
<oy>-0.94689514426939181</oy>
<oz>0.1643014214700198</oz>
<ax>-0.81986676842371453</ax>
<ay>-0.32152050640446078</ay>
<az>-0.47375420419862679</az>
<px>568.3780502864031</px>
<py>49.958773485311127</py>
<pz>353.43184246062265</pz>
<MotionType>FixedPose</MotionType>
<Pose>
<nx>0.50142183176690325</nx>
<ny>-0.0037616114188919747</ny>
<nz>-0.86519477397118882</nz>
<nw>0</nw>
<ox>0.27639578264144804</ox>
<oy>-0.94689514426939181</oy>
<oz>0.1643014214700198</oz>
<ow>0</ow>
<ax>-0.81986676842371453</ax>
<ay>-0.32152050640446078</ay>
<az>-0.47375420419862679</az>
<aw>0</aw>
<px>568.3780502864031</px>
<py>49.958773485311127</py>
<pz>353.43184246062265</pz>
<pw>1</pw>
</Pose>
<phase_error>
6.52313e-005 4.61277e-005 5.27545e-005 4.54155e-005 5.16256e-005 5.47379e-005 5.47153e-005 5.53385e-005 4.53949e-005 3.90839e-005 4.64643e-005 4.9345e-005 4.11487e-005 5.33012e-005 4.47133e-005 4.66567e-005 4.24385e-005 4.31861e-005 4.98001e-005 4.71818e-005 4.88509e-005 4.19001e-005 5.21342e-005 5.59035e-005 4.4167e-005 3.98714e-005 4.6836e-005 5.00563e-005 4.32968e-005 5.20853e-005 5.17326e-005 4.26449e-005 5.20655e-005 4.94252e-005 5.079e-005 5.29949e-005 5.42117e-005 5.34058e-005 4.63359e-005 4.82324e-005 5.03637e-005 4.50134e-005 5.6782e-005 5.48513e-005 5.06314e-005 4.83831e-005 4.82333e-005 4.55087e-005 3.70488e-005 5.31445e-005 4.99663e-005 4.5659e-005 4.13789e-005 4.18542e-005 4.24741e-005 4.53901e-005 3.56038e-005 4.36041e-005 5.23973e-005 5.14984e-005 5.49663e-005 5.37526e-005 4.66972e-005 4.83195e-005 4.88527e-005 5.05888e-005 5.01958e-005 5.41784e-005 4.01349e-005 5.1473e-005 4.99957e-005 5.18452e-005 4.64748e-005 5.58652e-005 4.04893e-005 5.35941e-005 5.31878e-005 5.77654e-005 4.97671e-005 5.04993e-005 5.35563e-005 4.39598e-005 4.49403e-005 4.77993e-005 4.05233e-005 4.86223e-005 4.6144e-005 4.5929e-005 4.70675e-005 4.99327e-005 4.66739e-005 4.55584e-005 4.99519e-005 4.57219e-005 5.48299e-005 4.5314e-005 4.57764e-005 4.87537e-005 4.51844e-005 5.4423e-005 4.62341e-005 4.91385e-005 5.07609e-005 5.45795e-005 4.81677e-005 4.40325e-005 5.33614e-005 5.55793e-005 5.42419e-005 4.23663e-005 4.35251e-005 4.1114e-005 5.0096e-005 4.85738e-005 4.65393e-005 5.83546e-005 4.55824e-005 5.51508e-005 5.20349e-005 4.06185e-005 4.3846e-005 4.60857e-005 3.7334e-005 3.68754e-005 4.62002e-005 4.29442e-005 3.83629e-005 4.10388e-005 4.29274e-005 4.41504e-005 5.47751e-005 4.124e-005 4.63963e-005 5.49523e-005 3.91168e-005 4.91823e-005 5.0899e-005 5.20812e-005 6.18092e-005 4.63392e-005 4.97727e-005 4.89398e-005 5.27795e-005 6.08669e-005 4.80167e-005 4.81886e-005 4.70479e-005 4.65286e-005 5.81741e-005 5.11303e-005 5.40692e-005 5.55856e-005 5.40139e-005 4.85012e-005 4.6211e-005 4.98757e-005 5.0152e-005 4.08173e-005 4.2325e-005 4.62668e-005 4.20222e-005 4.29153e-005 3.93754e-005 4.04137e-005 4.93021e-005 4.72691e-005 3.96624e-005 5.34673e-005 5.4235e-005 4.42795e-005 4.98188e-005 5.30178e-005 4.90756e-005 5.26742e-005 5.45939e-005 5.65092e-005 5.03312e-005 4.92253e-005 5.65847e-005 5.47409e-005 4.72891e-005 5.43883e-005 4.79189e-005 4.59807e-005 6.05179e-005 5.08367e-005 4.83954e-005 4.6113e-005 4.25287e-005 5.49439e-005 4.78683e-005 5.61996e-005 4.81826e-005 5.32562e-005 4.90282e-005 4.33851e-005 4.90001e-005 4.29398e-005 4.04782e-005 4.52788e-005 4.63935e-005 4.16226e-005 5.16106e-005 4.61232e-005 4.45048e-005 4.95204e-005 5.0152e-005 5.82078e-005 5.39653e-005 6.4679e-005 5.22614e-005 5.05301e-005 4.79772e-005 5.07749e-005 5.25254e-005 5.56555e-005 5.41784e-005 5.05801e-005 4.64848e-005 5.34619e-005 5.15562e-005 4.74206e-005 4.17609e-005 5.23076e-005 5.03637e-005 5.3103e-005 4.40006e-005 5.1398e-005 4.8788e-005 4.32548e-005 4.99004e-005 5.12029e-005 4.13869e-005 4.63136e-005 4.30743e-005 4.60307e-005 5.17445e-005 4.46267e-005 4.95911e-005 5.54162e-005 5.31396e-005 4.6061e-005 5.32076e-005 4.39598e-005 4.76259e-005 5.06221e-005 5.22733e-005 4.86374e-005 5.05621e-005 4.95099e-005 5.60414e-005 5.35132e-005 5.88716e-005 5.96183e-005 4.72918e-005 4.95305e-005
</phase_error>
vvs
 
Posts: 16
Joined: Sat Dec 19, 2015 3:42 am

Re: Incorrect alignment when fusing scans

Postby vvs » Mon Mar 21, 2016 7:26 pm

Some more info on my setup:
Camera: Canon EOS 550d (Live View via ESSDK)
Projector: ACER K132
David version 4.4
Calibrated at 60 mm using David structured light calibration pattern (http://www.david-3d.com/en/products/acc ... s/cp-set02)
vvs
 
Posts: 16
Joined: Sat Dec 19, 2015 3:42 am

Re: Incorrect alignment when fusing scans

Postby vvs » Mon Mar 21, 2016 8:33 pm

Also, clarifying what I meant by "poses" -- the model was simply rotated in place by about 7 - 8 degrees. It was not touched in any way (except for rotation - which was done via a turntable).
vvs
 
Posts: 16
Joined: Sat Dec 19, 2015 3:42 am

Re: Incorrect alignment when fusing scans

Postby micr0 » Tue Mar 22, 2016 10:19 pm

Make sure that the scans aren't distorted. This is usually what is happening when it seems that they aren't aligning. Distortion means that the scans are not be geometrically faithful to the model. This can be caused by a number of things.

µ
µ
micr0
 
Posts: 315
Joined: Wed Mar 19, 2014 3:54 pm
Location: NYC

Re: Incorrect alignment when fusing scans

Postby Sven » Wed Mar 23, 2016 4:42 pm

Hi vvs,

First of all, it seems you are aligning and fusing scans which have already been fused? You shouldn't do that. You should not import Fusion result and Fuse them again.
Instead please align the original scans, and after you have aligned all, Fuse them, Export, done.

Please look at the scans without texture, then you can see more easily how good the alignment is and whether they fit together. If they really don't fit, as micr0 said, they may be distorted. Usually the cause would be incorrect calibration (scanner configuration changed after last calibration) - in doubt, please calibrate again.

Sven
User avatar
Sven
DAVID Developer
 
Posts: 1541
Joined: Mon Jan 22, 2007 8:59 am
Location: Braunschweig, Germany

Re: Incorrect alignment when fusing scans

Postby vvs » Sat Mar 26, 2016 11:49 pm

Thanks Sven and micr0 for your help. Using your advice I've been able to get a much better result, but still not perfect.

Here is what I did to get a better result:
1) Increased angle between projector and camera to 25 degrees
2) Took 4 scans per pose with different exposure.

The result is my alignment and resulting topology looks better - here is the fused version. Unfortunately there are still a lot of ripples. I made sure they Keystone feature is turned off on my projector.
Image

The bigger issue though is the blending artifacts visible on the texture. Any ideas on what could be causing this?
Image

Thanks
vvs
 
Posts: 16
Joined: Sat Dec 19, 2015 3:42 am

Re: Incorrect alignment when fusing scans

Postby pottersfriend » Sun Mar 27, 2016 12:16 am

What is your camera set to? Is it BY8 or Y800? It almost looks like it might be set to Y800.
pottersfriend
 
Posts: 153
Joined: Wed Jun 03, 2009 1:49 am
Location: Washington State, USA

Re: Incorrect alignment when fusing scans

Postby vvs » Sun Mar 27, 2016 12:45 am

I'm using a Canon EOS 550d camera. Not using the David industrial camera.
vvs
 
Posts: 16
Joined: Sat Dec 19, 2015 3:42 am

Re: Incorrect alignment when fusing scans

Postby vvs » Sun Mar 27, 2016 12:50 am

From the David debug window, the Canon image grabber (ESSDK) format is RGB24.
vvs
 
Posts: 16
Joined: Sat Dec 19, 2015 3:42 am

Re: Incorrect alignment when fusing scans

Postby pottersfriend » Sun Mar 27, 2016 4:51 pm

Sorry, guess I should've read all the details.
pottersfriend
 
Posts: 153
Joined: Wed Jun 03, 2009 1:49 am
Location: Washington State, USA

Re: Incorrect alignment when fusing scans

Postby vvs » Mon Mar 28, 2016 2:16 am

I've tried pretty much everything and I simply cannot get a good texture fusion result. I'm happy with the merged mesh produced by ShapeFusion and made sure that the exposure of all of the scan's images is consistent, but no dice.

Not sure what algorithm Shape Fusion uses to combine textures, but this is a common problem in other domains (photogrammetry, panorama image stitching, etc). Does David use any blending algorithm like feathering or Gradient domain? Weird that I'm not able to get a good result here. Not sure if I am doing something wrong.

Here is a screenshot of the mesh untextured. Looks generally good - some minor ripples but that is fine.
Image

The merged texture looks quite bad. As I posted earlier when I do shape fusion of just one scan, the texture looks great. Its the merged texture when fusing multiple scans that is bad.
Image

@Sven, and anyone else from David, I have shared the scans and the fused result here.

Scans: https://drive.google.com/file/d/0B2aExW ... sp=sharing
Fused scan: https://drive.google.com/file/d/0B2aExW ... sp=sharing

Greatly appreciate any help. Thanks.
vvs
 
Posts: 16
Joined: Sat Dec 19, 2015 3:42 am

Re: Incorrect alignment when fusing scans

Postby micr0 » Mon Mar 28, 2016 5:17 am

I thought "generally fine" was as good as I was going to get until I actually got it right. That said, it seams (and I don't have an SLS-X kit) this is a bit of an art form. First start with your set up and make sure that your rig is not too close to the object. For a while I was placing the scanner equipment as close to the subject as possible and ultimately found that when the camera/projector were a little further back, the distortion that lead to bad alignments started to be reduced. Next we are assuming that you have a good calibration with an appropriate sized calibration pattern and that your camera angle to the projector is between 20° and 30°. Also make sure that the projector has the key stoning set to off and that if there are any aspect ratio or scaling adjustments in the projector that they too are off.. If your projector had a "pc" mode or presentation mode try that. My system scans better when my projector is set thusly.
µ
micr0
 
Posts: 315
Joined: Wed Mar 19, 2014 3:54 pm
Location: NYC

Re: Incorrect alignment when fusing scans

Postby vvs » Mon Mar 28, 2016 6:27 am

I was hoping this would be more of a science than an art :)

1) The distance between my scanning rig and the object is roughly 2 feet
2) The angle between the projector and scanner is roughly 28 degrees.
3) The horizontal distance between the projector and scanner is roughly 15 inches
4) I use the same projector as the SLS kit (Acer K132), so I know its a supported piece of equipment
5) Keystone correction is turned off in the projector
6) My calibration panel comes straight from David and the rig was calibrated at 60mm.

I can't understand why there would be such a discrepancy between the fused surface and the fused texture. It appears to me to be all about blending, which is why I asked about the blending algorithm.
vvs
 
Posts: 16
Joined: Sat Dec 19, 2015 3:42 am

Re: Incorrect alignment when fusing scans

Postby hunkatibor2 » Mon Mar 28, 2016 11:07 am

HI VHS
https://sketchfab.com/models/b97df32557 ... 0be8cda238
This is a textured examination made David their own hardware, the alignment algorithm works well .The learning curve is not the same in humans. Who fast who slow. You missed a lot of steps, I immediately started is the hardest. I wish you a very successful, I'm curious!
Tibor
Intel Xeon E5410 2.33GHZ 16GB RAM Windows 7 64bit David 4.5.0
2 Basler A622F camera 2 Tamron Asfirical AF 28-80mm lense
LG DLP Projector PF1500G Canon D40 Tokina 17-70mm 1:2.8-4,5
Humanti Calibration panel system and Alignmed&Two camera
hunkatibor2
 
Posts: 1408
Joined: Tue Mar 16, 2010 4:26 pm
Location: Hungary

Re: Incorrect alignment when fusing scans

Postby hunkatibor2 » Mon Mar 28, 2016 5:51 pm

Hi
I downloaded. Only David was able to open other programs do not.
First step reduced the mergers was 25%. The texture will not appear, only one white spot .This is not a good sign !! I note in the Annex, where is the big mistake. Clean the examination of these places. Reducing examining one by one down and then re-fusion.
Tibor
Attachments
2016-03-28_18-35-06.png
Intel Xeon E5410 2.33GHZ 16GB RAM Windows 7 64bit David 4.5.0
2 Basler A622F camera 2 Tamron Asfirical AF 28-80mm lense
LG DLP Projector PF1500G Canon D40 Tokina 17-70mm 1:2.8-4,5
Humanti Calibration panel system and Alignmed&Two camera
hunkatibor2
 
Posts: 1408
Joined: Tue Mar 16, 2010 4:26 pm
Location: Hungary

Next

Return to Structured Light Scanning

Who is online

Users browsing this forum: No registered users and 2 guests