Dear DAVID Community,

In 2016, DAVID and this forum have been integrated into HP Inc.
This forum has been read-only since then, and will be switched off by the end of 2019.

To everyone who has contributed here, we thank you for all your work and for the constructive, helpful and friendly atmosphere you have created!
Please continue to discuss on the HP Support Forums.

Alignment Problems

The place for all topics related to the 'Shapefusion' tool of DAVID.

Alignment Problems

Postby jharrison » Sun Jun 15, 2008 9:15 am

Hey everyone,

I seem to be having a problem when aligning my scans...

I have scanned 8 different views of an action figure, and everything seems cool during alignment, but when it comes to aligning the final scan, things just don't line up, though all of the surfaces look pretty aligned through the whole process.

I have tried different methods for aligning the scans, the 1st method was done by aligning each scan in a counter-clockwise manner starting with the front scan, but when it came to aligning the final scan (Front-Right) to the front scan there was a significant gap (Image 1 - circled is the right fist of the action figure)

Image

With the 2nd method I started with the front scan and worked around the model 180 degrees in each direction until reaching the back
scan, but I still received a gap of about the same size (Image 2 - circled is the base of the neck)

Image

I have tried both free rotation and rotation along the y-axis during alignment, but I still get the same results.

Any clues?

(Don't worry about the messy Geometry, this is just a test, and I plan on cleaning it up in Maya before the final alignment/fusion :) )

-J
User avatar
jharrison
 
Posts: 33
Joined: Tue Jun 03, 2008 1:34 am

Postby MagWeb » Sun Jun 15, 2008 11:01 am

Hi,

This may not be caused by Shapefusion, but by the precission of your scans. If your cam isn´t calibrated right and/or your corner isn´t precise, the meshes tend to be a little deformed (maybe a little flatter, little bowed in Y or something else) This little mistakes lead to the gab, finally visible if you try to align the last scan.

Suggestion to check: Scan your empty corner with switched off backgroundfilter ( Open config.xml in your DAVID folder with the win Editor and set the background filter factor to "-1" -deffault value 0,2). Now scan your planes and compare the 3D result with the grey grid of the 3D viewer. The result should exactly match to the grid - nothing should be bowed (also the result's edges)and the grid should be visible from both sides , even zoomed in.

My standard strategy for the align prozess:

Load all needed files (clockwise A B C D E F),
Arrange them;
Choose a mesh to start from (let's say A)
Hide C D E F
Align B to A
Hide A
Show C
Align C to B
Hide B
Show D
Align D to C
Hide C
Show E
Align E to D
Hide D
Now to close the gap with the last mesh:
Show A
Combine E and A
Show F
Align F to the combined E/A-mesh


Sometimes doing the fine registration twice improves the result.

Using the smoothing in the scanning step to much may cause altered topologie. I never use these buttons if I plan to fuse the meshes. Only interpolation is enough.


Hope that helps
Gunter
MagWeb
Moderator
 
Posts: 2499
Joined: Wed Jul 18, 2007 8:48 pm

Postby jharrison » Sun Jun 15, 2008 8:34 pm

Gunter,

Thank you for the timely response! It appears as if our alignment approach is the same, but when I get back to work (where my setup is), I will try the test to determine if there is a calibration issue.

I do use a right angle ruler to make sure the bottom, center, and top of the calibration corner are at 90 degrees, but maybe there is something I'm missing.

Thanks again for the advice, and I will reply with any found results!

Cheers,
-J
User avatar
jharrison
 
Posts: 33
Joined: Tue Jun 03, 2008 1:34 am

Postby MagWeb » Sun Jun 15, 2008 9:01 pm

Hi again,

Just a note to add:

In most cases the mistake isn´t the rightangled corner, but the fixing of the marker-patterns:
Their inner colums have to have a precise distance of 40mm from their centers to the origin (if you use the standard 60mm scale) AND
the horizontal rows at exactly the same height in Y.

Good luck

Gunter
MagWeb
Moderator
 
Posts: 2499
Joined: Wed Jul 18, 2007 8:48 pm

Postby jharrison » Mon Jun 16, 2008 9:03 am

The Calibration sheets I printed were from the Calibpoints_Scale60_DIN_A4_leftSide.pdf (the right as well). I didn't adjust any scale setting before printing, and when they printed, the scale measurement came out to 55mm (the print was auto-adjusting to the 8.5 X 11 paper)

I lined them up as instructed, so as expected, the distance from the center of the two inner columns should be 2/3 of the 55mm calibration scale distance? (roughly, 37mm) (- I'm guessing since you are informing me that the measurement from inner column centers with the standard 60mm scale would be 40mm).

(My setup is at work, but I printed the sheets at home to do a quick measurement test, and indeed the measurement from the center of the inner columns to the center line is 37mm.)


Now, here is where I am getting a little confused. You say that the measurement of the centers of the inner colums to the origin line need to be 40mm (we'll base this off the 60mm calib measurement to make the numbers easier), which should be the same distance as the centers of the circles along the Y, but when measuring the centers on the Y circles on the calib sheet, they amount to 60 mm which is obviously not the distance from the center of the inner columns to the origin.

Is there something I am missing?

Again, there could be a totally differen't problem with my calibration that is causing these alignment errors with my scans, so I will be sure to check with your initial reply regarding the background filter setting in the config.xml file when I get to work tomorrow.

Thanks again Gunter, your insight is very helpful!

Cheers,
-J
User avatar
jharrison
 
Posts: 33
Joined: Tue Jun 03, 2008 1:34 am

Postby MagWeb » Mon Jun 16, 2008 9:50 am

Again, for I lost the Re for a timeout:

Hi

Sorry one may get my explanation wrong:

When I wrote:

the horizontal rows at exactly the same height in Y


I was not refering to the value of 40mm.
I meant, the row at the left pattern has to have the same height as the one on the right plane.

A little mismatch or rotation while fixing the patterns on the boards cause exactly your problem.

I managed same problems using setting screws to place the patterns exactly and doing several empty scans.

See:
http://www.david-laserscanner.com/forum/viewtopic.php?t=398

If you put in a wrong scalevalue (60 instead of 55) DAVID simply thinks the object is bigger (60 instead of 55) with no influence to the quality or distortion.

Gunter
MagWeb
Moderator
 
Posts: 2499
Joined: Wed Jul 18, 2007 8:48 pm

Postby jharrison » Tue Jun 17, 2008 2:16 am

Hey Gunter,

Gotcha on the Horizontal and Vertical rows having the same height - everything looks good there.

I tried changing the background filter to -1 and check the planes in the 3D view, but for some reason the planes will not scan - at least not enough to try and determine if they are the issue. I get a few pixels here and there, but that's about it...

Are there any additional settings I need to adjust in order to scan the calib planes??

-J
User avatar
jharrison
 
Posts: 33
Joined: Tue Jun 03, 2008 1:34 am

Postby MagWeb » Tue Jun 17, 2008 6:28 am

Hi,

That's strange.

You only have to set that value to -1
Line 63 of the config.xml should be:

<BACKGROUND_FILTER_FACTOR>-1</BACKGROUND_FILTER_FACTOR>

Did you try to restart DAVID? (pretty simple idea, but I have no other)

Gunter
MagWeb
Moderator
 
Posts: 2499
Joined: Wed Jul 18, 2007 8:48 pm

Postby jharrison » Tue Jun 17, 2008 8:22 am

Yea, I noticed that after calibration it was setting the background filter to 0.2 even if I edited the config.xml file before opening David, so I changed the background filter to -1 after calibration, but still received nothing.

I did one last check on the calib sheets on my setup before I left work today and noticed that the height of the upper right circle (large black with white center) on the right sheet is 1mm higher than the upper left circle on the left sheet.

I am going to look into your setup, adjust the panels so they are 100% correct, and try a new scan- hopefully that works!

I'm looking forward to posting some results, so hopefully I make some progress soon.

Thank you for all of the assistance, Gunter!
User avatar
jharrison
 
Posts: 33
Joined: Tue Jun 03, 2008 1:34 am


Return to Shape Fusion

Who is online

Users browsing this forum: No registered users and 1 guest

cron