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

Disappointing behaviour of registration

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

Disappointing behaviour of registration

Postby AIMED » Wed Aug 10, 2016 11:08 pm

Hi Guys
I am struggling from five days but constantly David failed me. I love registration algorithm of David software but this time its failed me. Even same spot scan needed two or three times to register accurately. Any suggestions?? I am trying to align scan a to scan b. I put some clay balls to align accurately but its taking too long. Sometimes two scans align in first try but normally its taking hours. I couldn't understand why???

sc1.JPG
Logitech QuickCAM Pro 9000
Toshiba Satellite, 4GB RAM, 1GB AMD display
Panasonic Projector (model unknown)
AIMED
 
Posts: 67
Joined: Wed Aug 31, 2011 12:12 pm
Location: Islamabad

Re: Disappointing behaviour of registration

Postby drndadoo » Thu Aug 11, 2016 9:42 am

Parts with the clay are good aligned, but part with the hole is too far away, how is that possible? Is that green a) a single scan or a aligned group?
drndadoo
DAVID Support
 
Posts: 194
Joined: Tue Mar 24, 2015 3:09 pm

Re: Disappointing behaviour of registration

Postby AIMED » Thu Aug 11, 2016 12:21 pm

Thanks buddy
If I give you close picture you can check clay parts are not aligned as well. As those are small ones its seems aligned but it contains same difference as part. Working continously and facing same issue. :( :( :(
Logitech QuickCAM Pro 9000
Toshiba Satellite, 4GB RAM, 1GB AMD display
Panasonic Projector (model unknown)
AIMED
 
Posts: 67
Joined: Wed Aug 31, 2011 12:12 pm
Location: Islamabad

Re: Disappointing behaviour of registration

Postby micr0 » Thu Aug 11, 2016 3:12 pm

AIMED wrote:Thanks buddy
If I give you close picture you can check clay parts are not aligned as well. As those are small ones its seems aligned but it contains same difference as part. Working continously and facing same issue. :( :( :(


Yes. Please post close up images of the three circled areas.
At quick glance, I'm thinking the reason your scans are not aligning is not because of Davids algorithm, but because the scans don't match closely enough (they are slightly distorted).
µ
micr0
 
Posts: 315
Joined: Wed Mar 19, 2014 3:54 pm
Location: NYC

Re: Disappointing behaviour of registration

Postby drndadoo » Fri Aug 12, 2016 9:12 am

If alignment is bad it is mostly because of the distortion in scans. There are numerous reasons why that happens. Can you make a picture of your setup and calibration panel? You can make a test scan with something that is easy to align. The best would be with two objects in scene, so there should be no error.
drndadoo
DAVID Support
 
Posts: 194
Joined: Tue Mar 24, 2015 3:09 pm

Re: Disappointing behaviour of registration

Postby AIMED » Fri Aug 12, 2016 12:30 pm

I was thinking about it. But I tried to scan same area two times and both scans look almost more then 95% OK. I tried it several times by scaning same area multiple times but results remain same.

3.JPG


Another phenomenon is
I made copy of a scan, remove all area except clay balls from copy and align it to previous one. Its gone good. Now I am aligning the scan to its copy (copy contain only clay area) but its failed.
2.JPG


1.JPG


:( :( :(
Logitech QuickCAM Pro 9000
Toshiba Satellite, 4GB RAM, 1GB AMD display
Panasonic Projector (model unknown)
AIMED
 
Posts: 67
Joined: Wed Aug 31, 2011 12:12 pm
Location: Islamabad

Re: Disappointing behaviour of registration

Postby AIMED » Fri Aug 12, 2016 12:32 pm

micr0 wrote:
AIMED wrote:Thanks buddy
If I give you close picture you can check clay parts are not aligned as well. As those are small ones its seems aligned but it contains same difference as part. Working continously and facing same issue. :( :( :(


Yes. Please post close up images of the three circled areas.
At quick glance, I'm thinking the reason your scans are not aligning is not because of Davids algorithm, but because the scans don't match closely enough (they are slightly distorted).


After large number of tries, that area aligned so right now I cant give you close picture. :(
Logitech QuickCAM Pro 9000
Toshiba Satellite, 4GB RAM, 1GB AMD display
Panasonic Projector (model unknown)
AIMED
 
Posts: 67
Joined: Wed Aug 31, 2011 12:12 pm
Location: Islamabad

Re: Disappointing behaviour of registration

Postby AIMED » Fri Aug 12, 2016 12:37 pm

The only method is working right now is to align by mouse moves as close as possible then pairwise fine alignment is working. But its failing a number of times by same message that those are not neighbours then it align those all of sudden.
Logitech QuickCAM Pro 9000
Toshiba Satellite, 4GB RAM, 1GB AMD display
Panasonic Projector (model unknown)
AIMED
 
Posts: 67
Joined: Wed Aug 31, 2011 12:12 pm
Location: Islamabad

Re: Disappointing behaviour of registration

Postby AIMED » Fri Aug 12, 2016 12:52 pm

I moved the scan to the target and click pairwise fine alignment. See what happened. I am facing this first time other wise I used david to align and fuse scans so many times.

5.JPG
Moved by mouse


After pairwise fine alignment

4.JPG
After pairwise fine alignment


Another try

6.JPG
another try
Logitech QuickCAM Pro 9000
Toshiba Satellite, 4GB RAM, 1GB AMD display
Panasonic Projector (model unknown)
AIMED
 
Posts: 67
Joined: Wed Aug 31, 2011 12:12 pm
Location: Islamabad

Re: Disappointing behaviour of registration

Postby AIMED » Fri Aug 12, 2016 1:06 pm

After failed tries got success all of sudden. (pairwise fine alignment)

Image
Logitech QuickCAM Pro 9000
Toshiba Satellite, 4GB RAM, 1GB AMD display
Panasonic Projector (model unknown)
AIMED
 
Posts: 67
Joined: Wed Aug 31, 2011 12:12 pm
Location: Islamabad

Re: Disappointing behaviour of registration

Postby AIMED » Fri Aug 12, 2016 1:19 pm

Image
Logitech QuickCAM Pro 9000
Toshiba Satellite, 4GB RAM, 1GB AMD display
Panasonic Projector (model unknown)
AIMED
 
Posts: 67
Joined: Wed Aug 31, 2011 12:12 pm
Location: Islamabad

Re: Disappointing behaviour of registration

Postby Sören » Fri Aug 12, 2016 2:10 pm

TL;DR at the bottom.

###

As micr0 and drndadoo already mentioned, your scans may be distorted, which may make registration difficult or impossible. But even if your scans are not distorted in any way, there is another "problem" with your scans that make them difficult to align: The shape of your scans.

Let's ignore your clay balls for a second. So, what you're trying to do is to align two (almost) completely flat scans to each other. And as you already imagined, without any hints not even a human could align these scans correctly to each other, because there is no information about what belongs where. That's why you added the clay balls in first place. And that's great.

So, now with the clay balls, the task to align these scans is quite easy for a human. We instantly know that the clay balls are special markers that need to be aligned and then the rest of the scans are magically also perfectly aligned. The computer does not know that the clay balls are markers that need to be aligned. To the computer, the clay balls are simply parts of the scan. It does not even know that the balls exist, let alone their purpose.

What DAVID does, roughly speaking and in its basic form, is to try to find the pose where the two scans have maximum overlap. And now it tries tons of poses and finds that it can align the flat surface easily with the flat surface of the other scan. And the flat surface is huge, so there is a lot of overlap. And it maybe finds poses where the clay balls match each other but parts of the flat surface have no counterpart because they are simply out of bounds of the other scan. And the clay balls are small, so probably there is less overlap.

So, that's why you have difficulties aligning the scans. The flat surface just matches too good. But you have some possibilities to improve alignment.

First of all, with your available version of DAVID, you could try to increase the clay ball to flat surface ratio on your objects. The bigger the clay balls in your scans are, the bigger is their share of the surface. So more visible clay ball surface equals more "weight" in the registration, which results in the algorithm to give more attention to them.

Another option can be found in DAVID 5. In DAVID 5, there is a new feature called "Use Surface Features" - a checkbox that can be enabled to give more information to the registration algorithm. Using these surface features, the algorithm will automatically calculate and look for uncommon parts of the surface. These uncommon surface parts (in your case the clay balls) will get more attention by the registration algorithm and thus hopefully improve the final alignment. The downside of this is that it takes more time to calculate the alignment, because more information are calculated and fed to the registration. This is not a magic silver bullet, so it can not solve everything, but with some scans can definitely help a lot.

On the website, there is a trial version of DAVID 5 available, so you can import your raw scans there and try it out without having to buy it immediately. Although you can't save the results in the trial version, you can see if it would help.

###

TL;DR:

Probably the clay balls are too small with respect to the super flat surface. Try out larger clay balls or the "Use Surface Features" checkbox in DAVID 5 (trial version available on the website).
Sören
DAVID Developer
 
Posts: 4
Joined: Mon Aug 17, 2015 7:40 am

Re: Disappointing behaviour of registration

Postby AIMED » Fri Aug 12, 2016 2:22 pm

Thanks Soren
What I find with this project is, as I mentioned earlier, I make copies of all scans. Then open the copies and remove the cloud except the clay balls area. Then I align all clay balls scans with each other and those done fine.
Now I have all scans aligned but only with clay balls area and no flat area in scans.Now I am aligning scan4 (complete scan) with scan4_copy (only clay balls of scan 4). There is no flat areas in scan that can distrack david and not a single clay ball is identical of another. In this case those two scans must align quickly and nicely. I am doing this from sometime with success. In this project david failed me in this too.
Even it says there is no neighbouring areas there. Then all of a sudden it aligns it perfectly, why????
same scans must be align in no time, isnt it????
Logitech QuickCAM Pro 9000
Toshiba Satellite, 4GB RAM, 1GB AMD display
Panasonic Projector (model unknown)
AIMED
 
Posts: 67
Joined: Wed Aug 31, 2011 12:12 pm
Location: Islamabad

Re: Disappointing behaviour of registration

Postby micr0 » Fri Aug 12, 2016 2:54 pm

drndadoo wrote:If alignment is bad it is mostly because of the distortion in scans. There are numerous reasons why that happens. Can you make a picture of your setup and calibration panel? You can make a test scan with something that is easy to align. The best would be with two objects in scene, so there should be no error.


I find the easiest is to take two scans of the calibration panels right after calibration. Start with two scans of the panels with out moving them. Then do two at slightly different angles. That is a good place to start.
µ
micr0
 
Posts: 315
Joined: Wed Mar 19, 2014 3:54 pm
Location: NYC

Re: Disappointing behaviour of registration

Postby drndadoo » Mon Aug 15, 2016 8:57 am

Sören gave a good explanation. If scans are not distorted then you need to increase clay balls to flat surface ratio or you can put some big object in your scene (which is the same principle) as an side object and do your scans.
For example if you scan a ball, it will be hard to aligne scans. If you scan two balls in same scene it will be done very fast and precise.
drndadoo
DAVID Support
 
Posts: 194
Joined: Tue Mar 24, 2015 3:09 pm


Return to Structured Light Scanning

Who is online

Users browsing this forum: No registered users and 4 guests