Machine Vision

cancel
Showing results for 
Search instead for 
Did you mean: 

Pattern recognition and stitching images to reconstruct a panorama

H, 

 

I am taking around two hundreds snapshots of an object of 1m long and I would like to recreate a full panoramic image.

 

I can basically do it manually (it works fine) by imposing a identical offset in between all pictures but sometimes a small delay in the image acquisition is perturbing the assembly.

 

Therefore, I would prefer to use an algorithm to do it automatically and find the best offset for each snapshot.

 

I have seen that a development already exists although this is far to powerfull for my application (http://zone.ni.com/devzone/cda/epd/p/id/6497). However, it is developped for Labview 2010 and I have only LV2009. I do have Vision 2010. 

 

Could anyone comment on this and let me know if it would be adapted to my needs?

 

If so, could anyone convert that vi's in 2009 format or provide something equivalent?

 

Thanking you in advance.

 

Cris.

 

 

0 Kudos
Message 1 of 5
(3,903 Views)

I don't have LV 2009, but I've saved it back so hopefully it works for you.  Looks like a nice solution - whether it will work will probably depend on the content of your images, and whether there are distinct objects to match up.  I've had a similar problem in the past, but rather than finding particles, I've computed a cross-correlation as the measure of where to overlap images.  You could probably adapt this code to use a different alignment technique.

0 Kudos
Message 2 of 5
(3,892 Views)

Hello Cris,

 

Can you try the converted application and let us know if it fulfills your needs?

 

If there are any other questions, then please don't hesitate to let us know.

Kind Regards,
Thierry C - CLA, CTA - Senior R&D Engineer (Former Support Engineer) - National Instruments
If someone helped you, let them know. Mark as solved and/or give a kudo. 😉
0 Kudos
Message 3 of 5
(3,885 Views)

Dear GregS, 

 

Thanks a lot.

 

Before contacting the forum, I tried the cross correlation on matrices of pixels (grayscale) and double checked with Matlab. It was not giving any results. I had a peak at the center of the cross correlation matrix decaying in every diirection. Maybe I did something wrong...

 

I used yesterday the Match Pattern and it gives reliable and robust results. However the reconstitution of the full image is not optimal so far as I take only the mean distance between all successive images to reconstitue my "panorama". I hope the stitcher will improve a little bit this.

 

I will try it as soon I can find half a day to do it.

 

Thanks a lot!

 

Cris.  

0 Kudos
Message 4 of 5
(3,878 Views)

Dear Thierry, 

 

I will do so! Thanks a lot for your prompt response.

 

Best regards.

 

Cris.

0 Kudos
Message 5 of 5
(3,877 Views)