Labview results different than Vision assistant

[alias100]

Hello,

I am using a Sony XCL-5000 camera link camera with the NI PCIe-1427, LabVIEW 8.2 and Vision Assistant 8.0.  My application is aligning a small optical lens array to another feature.  I have tried a couple of methods that work on a stationary image in the vision assistant, but do not work as well with live images in labVIEW.

I first tried using the find circle function.  After converting the script into labVIEW it finds different

centers to the circles so there is a lot of jitter.  I use cross hairs to align the lens to the stationary object.  The crosshairs move around and turn on and off so much it makes it hard to align.

The other method I tried was using the find circle edge.  This technique seems to find some of the lens (circle), but creates a much larger offset circle making the center information useless.

I use the same parameters in the LabVIEW software as I used in the vision assistant script that worked.  I enclosed the image if anyone can give me some pointers that would be appreciated.  Thanks.

[muks]

If you find the labVIEW result any different from vision assistant,It might be because u need to define a distinct IMAQ destination(BUFFER).Otherwise it should be the same.The vaying of results may also be due to too much of ambience affecting the image.

[muks]

Check the aperture level u r operating at

[alias100]

I am going to try using match geometry next.  That works in vision assistant and I can determine which method works best and then try to investigate the changes in the live image applications.  I think it is a matter of variations in the actual images.  I need the most forgiving technique to find the pattern yet maintain accurate center location of geometry (circle).  Thanks for your ideas.

[muks]

---

I need the most forgiving technique to find the pattern yet maintain accurate center location of geometry (circle)

---

Just a tip.Using a n appropriate ROI for pattern recognition will  make your result  very sccurate.Like thresholding the image finding a circle by circle detection and defining an  roi based on that and feeding that area  as the search area will be helpful